Professional Documents
Culture Documents
Volume 7 2003
CONTENTS
Associate Editor
Serena Wieder, Ph. D.
Assistant Editor
Georgia DeGangi , Ph.
Treatment and Learning Center
Administrative Editor
Jo Raphael , M.
Editorial Board
Abstract. Long before children acquire an understanding of the semantic and syntactic char-
acteristics of spoken language, children have already developed a vast knowledge about the prag-
matics of human communication. To a large extent, this knowledge develops during preverbal
interactions between the infants and their parents. Emphasizing the continuity between prever-
bal and verbal communication is of great intuitive appeal; however, it has proven difcult to
emPirically validate specifc links between the infants ' preverbal experiences and the subsequent
origin of spoken language. The findings reviewed in this paper provide such a specifc link.
During the second half of the first year, tyPically develoPing infants and their parents start to
collaborate in establishing a shared interest in the world around them. Our findings suggest
that, at least in the case of autism, the success of the parent- child dyad in managing such a
shared interest in external objects or events is developmentally linked to children s long- term
language outcomes.
Learning a native language is an accomplishment that is within the grasp of most tod-
dlers. Yet, as Jerome Bruner and Virginia Sherwood put it discovering how children
do it has eluded generations of philosophers " (Bruner & Sherwood , 1983 , p. 41).
Early accounts ,stretching historically from St. Augustine to traditional learning the-
ory, made us believe that language acquisition is essentially a process of forming
associations between words on the one hand and agents , actions , and objects on the
other (O' Connell , 1969; Skinner , 1957). The limitations of this associationist view
became obvious in the late 1950s , particularly in the work of Noam Chomsky
(Chomsky, 1957). Chomsky s insight was that the process of language acquisition is
far more complex than connecting words with meaning. Children also acquire a
grammar - an underlying system of rules enabling them to generate an infinite num-
ber of new sentences. Importantly, in shifting the emphasis from word learning to
rule acquisition , traditional principles of associative learning suddenly appeared il
equipped to explain children s growth in language. It seemed unreasonable to
assume that children could learn about the underlying structure of language just by
imitating their elders and exposure to systematic reinforcement. Given the con-
strained view on learning predominant at the time , Chomsky took his argument as
far as to question the learnability of language altogether. He argued that all human
001- 018 10/28/03 12: 19 PM Page 2
177235-
languages share a common set of syntactic rules - these rules are part of children
innate endowment , ready to be recognized as soon as children are exposed to their
native languages. It wasn t until the early 1980s that Bruner pointed out that parents
participate in language learning in a manner far beyond serving either as a model to
imitate or as a source of reinforcement (Bruner , 1981; Bruner & Sherwood , 1983).
From early on , mother and child develop conventions for carrying out joint tasks. It
is the structure inherent in these joint tasks that may shape children s learning about
the structure underlying spoken language.
Bruner s proposal of developmental continuity between preverbal and verbal
exchanges was sparked by John Austin s insight that, even when mature language is
considered , word meaning and grammar are insufficient to convey a speaker s com-
municative intent (Austin , 1962). Bruner provides the example of a trivial question
that is being asked at a dinner party: " Would you be so kind as to pass the salt?" Very
likely, the listener wil interpret this question as a polite request for the salt rather
than an inquiry concerning his or her perceived kindness. This interpretation is pos-
sible a) because of the listener s familiarity with the communicative context , and b)
because the listener shares some of the speaker s notions about intentions. For devel-
opmental psychologists , this characterization of mature language use is intriguing
because it points in the direction where continuity may be found between preverbal
and verbal exchanges. Long before children appreciate the semantic and syntactic
characteristics of spoken language , they already have acquired a rich body of knowl-
edge about the context in which communication takes place as well as intentional
states that motivate people to communicate. This knowledge may serve as a " matrix
into which syntactic and semantic achievements can be set" (Bruner , 1981 , p.l59).
To a large extent, Bruner s empirical work on the origin of language was con-
cerned with describing and characterizing the early social experiences of infants.
Based on a number of extensive longitudinal single case studies , Bruner emphasized
that the communicative life of preverbal infants is limited to a rather small number
of familiar interactive routines (Bruner & Sherwood , 1983; Ninio & Bruner , 1978;
Ratner & Bruner , 1978). In his view , these familiar routines provide the infant with
the experiences necessary to interpret the communicative intentions of other people:
Infants and their mothers develop and conventionalize the rule structure governing
the exchange , infants learn to make their intentions plain , and mothers learn to inter-
pret the intentions of their children. Moreover , Bruner argues that the vast majority
of early mother- infant exchanges are governed by three basic communicative inten-
tions: affiliating, requesting, and sharing.
During the first six months of life , most social encounters between mother and
infant are aimed at establishing affiliation. Initially, interactions are concerned with
the mutual adaptation between the infant's cycling physiological needs and the
mother s attempts to anticipate and fulfill these needs. For example , studying early
feeding interactions , Kenneth Kaye was able to show that even during the first two
weeks of life mothers and infants learn to read each other s signals and to anticipate
one another s behaviors (Kaye , 1977 , 1982). As infants achieve control over their own
gaze during the second month (Bronson , 1974), the face- to- face encounter becomes
, &
the primary context for social interaction. But again , interactions are directed
towards establishing affiliation: Mothers and infants learn to regulate eye contact and
arousal , to share each other s emotions , and to respond to each other s facial cues
(Brazelton , Koslowski , & Main , 1974; Cohn & Tronick , 1988; Kaye & Fogel , 1980;
Stern , 1974).
As infants move through the second half of the first year , mothers and infants
increasingly communicate about objects and events outside the dyad (e. g. toys or
food). In managing such a triadic setting, involving mother , infant , and a toy, two
communicative intentions become visible. First , mother and infant develop interac-
tive routines aimed at regulating each other s object directed behaviors. Initially,
these routines have the exchange of objects as their common theme. The infants
learn to request objects being held by their mothers and also to return them when
being prompted to do so. Later , the interactive routines increase in sophistication
now being directed at obtaining help in carrying out one s own goal directed actions
(e. g. blowing up a balloon). The second communicative intention emerging during
the second half of the first year is directed towards achieving and maintaining a
shared interest in surrounding objects and events. That is , mothers and infants
develop interactive routines concerned with directing and following each other
attention. Here, communication is not instrumental in the sense that it is intended to
obtain objects or services from another person. Rather , it serves the sole purpose of
recruiting another person s attention to objects , states , and topics that are at the cen-
ter of one s own attention.
Empirical evidence linking infants ' experiences within these early interactive rou-
tines to children s subsequent rate of language acquisition has been scarce. Rare
exceptions are a number of longitudinal studies that demonstrate the predictive sig-
nificance of the dyad' s success in managing a shared interest in surrounding objects
or events. In several longitudinal studies on play interactions between 9 to 15 month
old infants and their mothers , Melinda Carpenter , Michael Tomasello , and col-
leagues have demonstrated that infants who spent more time in a state of shared toy-
engagement with their mothers subsequently developed language at a faster rate than
infants who spent less time in shared toy- engagement with their mothers initially
(Carpenter , Nagell , & Tomasello , 1998; Tomasello , MannIe , & Kruger , 1986;
Tomasello & Todd , 1983). In addition , researchers have also studied the individual
contributions of mothers and infants in establishing such a shared interest in the
external world. Focusing on the parents ' interactive behaviors , several longitudinal
studies have shown that the tendency to describe objects or events occupying their
children s current focus of attention reliably predicted the rate of children s subse-
quent language learning. (Akhtar , Dunham , & Dunham , 1991; Bornstein & Tamis-
LeMonda , 1989; Dunham & Dunham , 1992; Smith , Adamson , & Bakeman , 1988;
Tamis- LeMonda , Bornstein , & Baumwell , 2001; Tamis- LeMonda , Bornstein
Damast, 1996; Tomasello & Farrar , 1986). Similarly, with respect to the infants ' early
communication skils , Jessica Markus and colleagues have shown that the twelve
month olds ' capacity to follow another person s pointing gesture to an interesting
001- 018 10/28/03 12: 19 PM Page 4
177235-
gle mother- infant dyad , Trevarthen & Hubley (1977) described that it was in the con-
text of such a game that the infant for the first time attended to the mother s face while
playing with an object. Using a larger longitudinal sample of 28 mother- infant dyads
Bakeman and Adamson were able to provide further documentation of this maternal
tendency to tune into the infants ' ongoing engagement with objects (Adamson &
Bakeman , 1984 , 1985; Bakeman & Adamson , 1984 , 1986). Importantly, even though
at first infants may show no explicit acknowledgement of their mothers ' attempt to
join their ongoing toy engagement (for example by alternating their gaze between the
toy and their mothers ' face), findings from this longitudinal study suggest that these
episodes of supported joint engagement provide an important context for the infants
social learning. First, the authors reported that the infants ' earliest referential gestures
(e. g. pointing) emerged in situations where parents had tuned into their infants ongo-
ing engagement with a toy. Second , Bakeman and Adamson were able to show that
episodes characterized by a shared interest in an object are generally accompanied by
elevated levels of affective involvement on behalf of the infant. Similar levels of affec-
tive involvement are rarely observed when children played with the toys on their
own. Interestingly, infants showed the same elevated levels of affective involvement
during joint engagement, independently of whether they explicitly acknowledged
their mothers ' presence or not.
In summary, by joining the infants ' ongoing toy engagement , parents assure the
infants ' participation in social encounters that involve a shared interest in an object
or event. Frequent participation in such encounters provides the infant with the expe-
riences necessary to acquire the social skils needed to playa more active role in
managing such encounters. Around nine months of age , a range of such social behav-
iors emerge in the behavioral repertoire of typically developing infants: infants start
to follow their mothers ' gaze direction and pointing gesture in order to determine the
target of mom s interest, infants alternate their gaze between an interesting object or
event and their mothers ' face in order to check ' if she also sees , and they start to
point to and show objects in order to capture their mother s interest. This group of
behaviors is frequently referred to as joint attention behaviors (Tomasello , 1999;
Tomasello , Kruger , & Ratner , 1999).
each other at a small table. A set of toys is in view but out of reach of the child. The
tester presents a variety of toys and plays a series of games aimed at eliciting request-
ing, sharing, and affiliating behaviors. In addition , on several trials , the tester points
to the left, right , and behind the child to evaluate whether the child follows the exper-
imenter s bids by turning his or her head in the designated direction. The procedure
is videotaped and coded for the frequency of specific social behaviors. These behav-
iors fall into five mutually exclusive categories: a) initiates social interaction (affliation):
g. initiates a turn taking game by rolling a ball towards the experimenter; b) responds
to bids for social interaction (affliation): g. establishes eye contact with an adult dur-
ing a pause in a face- to- face game , or returns the ball to the experimenter during a
turn taking game; c) initiates joint attention: (sharing): g. alternates gaze between an
active mechanical toy and the experimenter s face , establishes eye contact while
examining a toy, points to toys within reach , or shows an object to the experimenter;
d) responds to bids ftr joint attention (sharing): g. follows the experimenter s pointing
gesture towards a poster on the wall; e) initiates behavior regulation (requesting):
reaches towards a toy that is out of reach , establishes eye contact with the experi-
menter when a mechanical toy ceases , or points to a toy out of reach. Interrater reli-
ability for this scale has been established multiple times and mean generalizability
coefficients are at about . 80.
Findings from this research suggest that children with autism show deficits across
a range of joint attention behaviors. As compared to language matched controls , chil-
dren with autism were less likely to respond to the experimenter s pointing gesture;
they were less likely to initiate joint attention by alternating gaze between a toy and
the experimenter s eyes; and they were less likely to initiate a shared interest by
pointing at a toy or by showing a toy to the experimenter or a parent. In contrast, a
mixed picture of deficits and strengths emerged with respect to requesting and social
interaction (affiliating) behaviors. Although children with autism were found to initi-
ate social interaction less than children with Down syndrome , no group differences
were found between children with autism , typically developing children , and chil-
dren with heterogeneous developmental delay. Similarly, children with autism did
not differ from matched controls in their responsiveness to social interaction. Finally,
with respect to requesting behaviors , we found that typically developing children
showed on average more requesting behaviors than any of the other three groups. In
turn , a deficit in requesting behaviors is characteristic not only for children with
autism but also for children with other developmental delays. However , we also
found evidence that the deficit in requesting behaviors is somewhat more profound
in children with autism than it is in children with heterogeneous developmental
delays.
In reviewing the literature on typically developing infants , we emphasized that a
shared interest between parent and infant emerges based on contributions from both
partners. We also emphasized that especially during early stages of development the
responsibilities for assuring a shared interest are not distributed evenly between par-
ent and infant. Early in development, when infants stil lack the skils necessary to
play an active role in managing a shared interest in an object , parents respond to the
001- 018 10/28/03 12: 19 PM Page 7
177235-
infants ' limitations by adopting a tendency to tune into their children s ongoing toy
engagement. Given that children with autism also lack the skils necessary to manage
a shared interest with their parents , the question arises , how do parents of children
with autism cope with their children s behavioral limitations?
The few studies that have compared the interactive behaviors of parents of chil-
dren with autism to the behaviors of parents of typically developing children have
documented that parents of children with autism tend to be more directive and
demanding (Kasari , Sigman , Mundy, & Yirmiya , 1988; Watson , 1998). That is , they
tend to initiate more activities , show more physical prompts , and use more redirect-
ing utterances. Similar findings are also often reported when behaviors of parents of
developmentally delayed children are compared to parents of typically developing
children (Cunningham , Reuler , Blackwell , & Deck , 1981; Eheart , 1982;Jones , 1977).
At this point , very little is known as to why parents cope with the behaviorallimita-
tions of young children with autism by assuming a more directive role during play
interactions. Part of the reason may be that the spontaneous play behaviors of young
children with autism are often rather narrow in content, involve many repetitions , and
lack functional or symbolic play acts. In communicating clear expectations about what
the child should do and how the child should do it, parents may try to advance their
children s object directed play behaviors. Despite the good reasons that parents may
have for assuming a more directive and demanding role during interactions , such
interactive behaviors may compromise an important goal of parent- child interaction:
the mutual managing of a shared interest in surrounding objects or events. As empha-
sized above , parents of typically developing infants initially assure a shared interest by
joining the infants ' spontaneous object engagement. They do so , even though at first
the infant might not even acknowledge the parent's involvement. It might be that the
emerging social understanding and language development of young children with
autism would benefit from the same kinds of interactive experiences.
mean chronological age = 12 years 10 months), and once again during early adult-
hood. Data from the early adulthood follow-up are stil in the process of analysis. For
this reason , findings from this second follow- up wil only be reported for a sub- sam-
ple of 19 early adults (mean chronological age = 20 years 3 month), and only as part
of a longitudinal follow-up for a sub- study on early mother- child play interaction.
The initial diagnosis of each child was made independently of the research by a
group of clinicians experienced in the evaluation of developmental disorders using
DSM III (American Psychiatric Association , 1980) criteria. Initial clinical diagnoses
were supplemented with diagnoses based on two other sources , a videotaped inter-
."' ,&
001- 018 10/28/03 12: 19 PM Page 8
177235-
action scored with the Childhood Autism Rating Scale (CARS , Schopler , Reichler
Renner , 1986), and an interview during which the Autism Behavior Checklist (ABC
Krug, Arick , & Almond , 1980) was administered to the parents. As part of the follow-
up assessments , the diagnoses were confirmed using the Autism Diagnostic
Interview- Revised (ADI- , Ie Couteur , Rutter , Lord , & Rios , 1989; Catherine Lord
Rutter, & Le Couteur , 1994). At each assessment point, a standardized language test
was administered. The language measures used were either the Reynell
Developmental Language Scales (Reynell , 1977) or the Childhood Evaluation of
Language Fundamentals- Revised (CELF- , Semel , Wiig, & Secord , 1987), depend-
ing on children s skilleve1. Both measures provide age equivalence scores for chil-
dren s language abilities.
Figure 1 displays the individual language profiles of the participants in our study
as a function of chronological age. Only data from the initial and mid- school follow-
10
9 -
8 -
7 -
6 -
5 -
4 -
3 -
2 -
1 In two cases , language skils during early adulthood were assessed using the Mullen Scales of Early
Learning (Mullen , 1995) or the Childhood Evaluation of Language Fundamentals- Preschool (CELF-
Preschool , Wiig, Secord , & Semel , 1992).
001- 018 10/28/03 12: 19 PM Page 9
177235-
tion , initiates behavior regulation) were used to predict children s language scores
during the mid- school period. Given the stability of children s language abilities over
time , all analyses statistically controlled for individual variations in initial language
scores.
The results showed that children s responsiveness to bids for joint attention reli-
ably predicted their expressive language scores during the mid- school assessment;
r(33) = .46; Ol. That is , children who were more likely to respond to the experi-
menter s bids for joint attention during preschool had significantly better long- term
language outcomes than children who were less responsive initially. This predictive
relation was not explained by variations in initial language abilities. Besides joint
attention , none of the other measures of nonverbal communication showed a reliable
association with children s language outcomes. Neither their capacity to initiate social
interaction , nor their responsiveness to social interaction initiated by the experi-
menter reliably predicted children s mid- school language abilities. Further , no reli-
able association was found between the frequency of the preschoolers ' requesting
behaviors and language abilities during mid- school.
Predicting language growth from the parents ' interactive
engagement during play. During
the initial assessment of our longitudinal project, parents were provided with a stan-
dardized set of toys and asked to play with their children ' as they normally would'
The interactions were videotaped and later analyzed using a micro- analytic coding
system for which good inter- coder reliability has been demonstrated. In coding the
videotaped interactions , we focused on a range of interactive behaviors used by the
mothers and were particularly interested in how these behaviors related to children
ongoing focus of attention as well as children s ongoing object engagement. A com-
prehensive description of the coding system including a detailed report on the find-
ings is provided elsewhere (Siler & Sigman , 2002). For the purpose of this
presentation , we wil only focus on the parents ' verbal engagement during play. In a
first step, we coded the onset of each of the mothers ' verbal utterances. In a second
step, for each utterance it was decided a) whether the maternal utterance referred to
a toy that the child was already looking at, and b) whether the maternal utterance
reflected an intention to comment on the object or whether the utterance was
demanding the performance of a specific object directed behavior. The measure
included in our analysis was based on the frequency of maternal utterances that were
responsive to children s focus of attention and also reflected an intention to comment
rather than demand. This measure was referred to as a measure of maternal syn-
chronization.
For the data analysis , the primary operation was to use the measurements of
maternal synchronization to predict children s subsequent growth in language abili-
ties until mid- school and early adulthood. Findings from this analysis indicate that
'The final measure of maternal synchronization used in this analysis was based on the frequency of syn-
chronized maternal utterances but also controlled for a) the mothers ' overall verbal engagement , and b)
the duration children spent attending to toys. Neither of these control variables was in itself related to chil-
dren s language outcomes.
~, ." ~.
001- 018 10/28/03 12: 19 PM Page 11
177235-
mothers who showed higher levels of behavioral synchronization initially had chil-
dren who developed superior language skils until mid- school (r (17) = . 005)
and early adulthood (r (19) = . 001). Importantly, further analysis showed that
this predictive relation between maternal synchronization and children s subsequent
language growth was not due to variations in initial child characteristics such as lan-
guage , IQ, or joint attention.
Conclusion
Long before children acquire an understanding of the semantic and syntactic
characteristics of spoken language , children have already developed a vast knowl-
edge about the pragmatics of human communication. To a large extent , this knowl-
edge develops during preverbal interactions between the infants and their parents.
Emphasizing the continuity between preverbal and verbal communication is of great
intuitive appeal; however , it has proven difficult to empirically validate specific links
between the infants ' preverbal experiences and the subsequent origin of spoken lan-
guage. The findings reviewed above provide such a specific link. During the second
half of the first year , typically developing infants and their parents start to collabo-
rate in establishing a shared interest in the world around them. Our findings suggest
that , at least in the case of autism , the success of the parent- child dyad in managing
such a shared interest in external objects or events is developmentally linked to chil-
dren s long- term language outcomes.
Specifically, the research reviewed above had three major findings. First, when
compared to matched controls , children with autism showed a specific deficit in
behaviors used to establish joint attention with other people. Second , preschool aged
children with autism who were more responsive to another person s bid for joint
attention had reliably better language skils when assessed during mid- school than
did children who were less responsive to bids for joint attention initially. Third , par-
ents who were more responsive to their children s focus of attention and ongoing
activity during initial play interactions had children who developed language skils
until mid- school and early adulthood that were superior to children of parents who
were less responsive initially. Importantly, both predictive relations were independ-
ent from each other and could not be explained by variations in children s initial lan-
guage ability or overall intellectual level.
Until recently, children with autism were believed to suffer from a rather generic
social deficit, affecting all areas of human communication. According to the Third
Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM III), pub-
lished as late as 1980 , the social deficits of children with autism include a " failure to
develop interpersonal relationships a lack of responsiveness to and a lack of inter-
est in other people , a " failure to develop normal attachment behaviors , as well as
an " indifference or aversion to affection and physical contact" . During the last two
decades , researchers and clinicians became increasingly aware that the aloof picture
painted in these early accounts on autism did not do justice to the quality of the rela-
001- 018 10/28/03 12: 19 PM Page 12
177235-
tionships that children with autism entertain with other people. Much of this reeval-
uation was sparked by Bruner s insight that human communication serves a variety
of different functions. Recent research , evaluating the social characteristics of chil-
dren with autism across different communicative situations has identified a complex
pattern of strengths and weaknesses.
Social encounters between young children with autism and their parents are most
successful when directed at affiliation. Such exchanges frequently utilize a face- to-
face situation and involve activities such as tickling, chasing, rough and tumble play,
or peek- boo. In the context of these social activities , children with autism usually
require only minor support from their parents , participate with enjoyment and eye
contact , take turns and even take the initiative in order for the game to continue.
Along the same lines , a number of studies have shown that children with autism
once reunited with their parents after a few minutes of separation , demonstrate clear
signs of attachment (Capps , Sigman , & Mundy, 1994; Rogers , Ozonoff, & Maslin-
Cole , 1991 , 1993; Shapiro , Sherman , Calamari , & Koch , 1987; Sigman & Ungerer
1984). That is , during reunion episodes , children with autism were shown to direct
more social behaviors and physical contact to their parents than they do to a stranger.
Another social encounter of relative strength involves situations where children with
autism are motivated to request objects or assistance from another person. Clearly,
children with autism share the notion that other people can be instrumental for
acquiring help. This is not to say that the communicative means children with autism
use to communicate requests are as advanced as one would hope. In requesting
objects or assistance , children with autism tend to be limited to direct means of com-
munication (e. g. guiding another person s hand , reaching), lacking symbolic behav-
iors such as gestures and language.
Our research showed that the social difficulties of children with autism are most
apparent during social exchanges directed at establishing and maintaining a shared
interest in the surrounding world. This finding is very robust and has been replicated
in other research laboratories (Loveland & Landry, 1986). Also , most current diag-
nostic inventories consider joint attention behaviors as an important indicator for
autism , for example , the Autism Diagnostic Observation Schedule - Generic
ADOS- G (DiLavore , Lord , & Rutter , 1995), or the Modified Checklist for Autism in
Toddlers , M- CHAT (Charm an et al. , 2001; Robins , Fein , Barton , & Green , 2001).
Moreover , the success of the parent- child dyad in establishing a shared interest in
external objects , events , or topics appears to be linked to children s subsequent ori-
gin of spoken language. That is , both children s responsiveness to the parents ' bids
for joint attention and the parents ' responsiveness to the children s focus of attention
and ongoing activity reliably predicted the rate of children s subsequent language
acquisition.
It is important to emphasize that the predictive relation between preschoolers
access to certain interactive experiences Uoint attention , maternal synchronization)
and children s subsequent language development is correlational in nature.
Correlational findings do not allow us to draw firm conclusions about the underlying
causal mechanisms. That is , correlated variables , even when separated by time , do
,&
001- 018 10/28/03 12: 19 PM Page 13
177235-
not necessarily reflect a direct causal link but can also be the outcome of an uniden-
tified third variable. For example , it would have been possible that children s initial
language or overall intellectual functioning was linked to both , children s access to
interactive experiences and their language outcomes. Using methods of statistical
control , our data showed that both variables , initial language and IQ, were not able
to account for the predictive relations identified in our research. However , in longi-
tudinal research there is no way of knowing whether other variables that have not
been evaluated as part of the study could provide such an alternate explanation. In
order to identify the causal network underlying the predictive relations identified in
our research , future studies would need to move towards research designs where sub-
jects are randomly assigned to different treatment conditions.
According to Bruner , typically developing infants and their mothers collaborate
in developing interactive routines that enable infants to conventionalize three funda-
mental communicative intentions: affiliating, requesting, and sharing. In comparison
interactive exchanges between preschoolers with autism and their mothers are less
routinized and the underlying communicative intentions are much harder to detect.
Nevertheless , interventions targeting communication skils in young children with
autism have developed a variety of interactive strategies that help children with
autism to make their intentions plain. Many such strategies have focused on chil-
dren s requesting behaviors , for example: Incidental Teaching (McGee , Morrier
Daly, 1999), Pivotal Response Training (Koegel , Koegel , Harrower , & Carter , 1999;
Koegel, Koegel , Shoshan , & McNerney, 1999; Laski , Charlop, & Schreibman , 1988;
Schreibman , Kaneko , & Koegel , 1991), or Developmental , Individual- Difference
Relationship- Based Program , DIR , (Greenspan & Wieder , 1998 , 1999). Generally,
these approaches use environmental arrangements (e. g. placing attractive objects in
children s view but out of their reach) and interactive strategies (e. g. gently inter-
rupting the children s play by placing a hand on the toy, providing the children with
a choice of attractive items) to elicit a state of heightened emotions where children
are likely to request objects or assistance. Once children communicate this intent
(e. g. by guiding the adults hand towards the desired objects) the adult uses prompt-
ing strategies to shape children s communicative means (e. g. by purposefully misun-
derstanding the child' s request). Similar interactive strategies have been developed to
conventionalize children s affiliating behaviors. Intervention programs such as DIR
face exchanges to promote eye contact , to promote children s active
utilize face- to-
participation during reciprocal games , and above all to increase the length of the
communicative exchange (the number of circles of communication that are opened
and closed).
Developing interactive strategies that elicit and conventionalize acts of sharing
from children with autism has proven to be much more difficult. Our data suggest
that following children s lead and commenting on their ongoing experiences with
objects might be a first step. Joining the children in manipulating the preferred toy
might be a second step. However , to be successful in joining the object engagement
of children with autism , adults have to be very careful not to throw the child off by
being too demanding or intrusive. Encounters where adults and children are engaged
), ),
in parallel activities might be more successful if adults assume an attitude where chil-
dren are considered to be the expert on how to play with the toys. Once parallel
activities are established , the encounter can be elaborated in many directions: adults
may start to imitate some of their children s actions; they might develop games where
adults act contingent upon children s object directed behaviors; they might leave
some of their actions unfinished; or they might create situations where unexpected
things can happen. All these elaborations might prompt the children to acknowledge
the adults ' involvement by alternating their gaze between the toy and the adults ' face
or by completing or imitating the adults ' actions.
The account presented above emphasizes that humans communicate for various
reasons and that researchers or clinicians concerned with the development of com-
munication skils are well advised to carefully consider the different functions that
communication plays in the life of young children. Our research shows that this per-
spective is very productive not only for pinpointing the social deficits of young chil-
dren with autism but also for identifying social experiences that are linked to
children s subsequent language development. Particularly, our research emphasizes
that episodes of shared interest in an object , event, or topic provide a unique context
for language learning. Helping parents to be successful at managing such encounters
might be an important piece in a comprehensive intervention program targeting
communication and language in young children with autism.
References
Adamson , L. B. , & Bakeman , R (1984). Mothers ' communicative acts: Changes during
infancy. Infant Behavior Development 7(4), 467- 478.
Adamson , L. B. , & Bakeman , R (1985). Affect and attention: Infants observed with mothers
and peers. Child Development 56(3), 582- 593.
Akhtar , N. , Dunham , F. , & Dunham , P. J. (1991). Directive interactions and early vocabulary
development: The role of joint attentional focus. Journal of Child Language 18(1), 41-49.
American Psychiatric Association. (1980). Diagnostic and statistical manual ftr mental disorders
(3rd ed. ). Washington , DC , US: Author.
American Psychiatric Association. (1994). Diagnostic and statistical manual ftr mental disorders
(4th ed. ). Washington , DC , US: Author.
Austin J. L. (1962). How to do things with words. Oxford , UK: Clarendon Press.
Bakeman , R , & Adamson , L. B. (1984). Coordinating attention to people and objects in
mother- infant and peer- infant interaction. Child Development 55(4), 1278- 1289.
Bakeman , R , & Adamson , L. B. (1986). Infants ' conventionalized acts: Gestures and words
with mothers and peers. Infant Behavior Development 9(2), 215-230.
Bornstein , M. H. , & Tamis- LeMonda, C. S. (1989). Maternal responsiveness and cognitive
development in children. In M. H. Bornstein (Ed. Maternal responsiveness: Characteristics
and consequences (Vol. 43 , pp. 49- 61). San Francisco , CA , US: Jossey- Bass.
Brazelton , T. B. , Koslowski , B. , & Main , M. (1974). The origins of reciprocity: The early
mother- infant interaction. In M. Lewis & L. A. Rosenblum (Eds. The efct of the infant
on its caregiver (pp. 49- 76). New York , NY , US: Wiley.
Bronson , G. (1974). The postnatal growth of visual capacity. Child Development 45(4), 873- 890.
Bruner , J. (1981). The social context of language acquisition. Language Communication 1 (2-
sup- 3), 155- 178.
),
Bruner j., & Sherwood , V. (1983). Thought, language , and interaction in infancy. Inj. D. Call
E. Galeson & R L. Tyson (Eds. Frontiers of infant psychiatry (pp. 38-52). New York , NY
US: Basic Books.
Capps , L. , Sigman , M. , & Mundy, P. (1994). Attachment security in children with autism.
Development Psychopathology, 6 (2), 249-261.
Carpenter , M. , Nagell , K , & Tomasello , M. (1998). Social cognition , joint attention , and com-
municative competence from 9 to 15 months of age. Monographs of the Society for Research
in Child Development 63(4), 176.
Charman , T. , Baron- Cohen , S. , Baird , G. , Cox , A. , Wheelwright, S. , Swettenham , j., et al.
(2001). Commentary: The Modified Checklist for Autism in Toddlers. Journal of Autism
and Developmental Disorders 31(2), 145- 148.
Chomsky, N. (1957). Syntactic structures. The Hague , The Netherlands: Mouton.
Cohn , j. F. , & Tronick , E. Z. (1988). Mother- infant face- to- face interaction: Influence is bidi-
rectional and unrelated to periodic cycles in either partner s behavior. Developmental
24(3), 386-392.
Psychology,
Cole , M. John- Steiner , v. , Scribner , S. , & Souberman , E. (1978). Mind in society: The development
of higher psychological processes. L. S. Vygotsky. Cambridge , MA , US: Harvard University
Press.
Collis , G. M. , & Schaffer , H. R (1975). Synchronization of visual attention in mother- infant
pairs. Journal of Child Psychology Psychiatry Allied Disciplines 16 (4), 315-320.
Cunningham , C. E. , Reuler , E. , Blackwell j., & Deck j. (1981). Behavioral and linguistic devel-
opments in the interactions of normal and retarded children with their mothers. Child
Development 52(1), 62- 70.
DiLavore , P. C. , Lord , C. , & Rutter , M. (1995). Pre- linguistic autism diagnostic observation
schedule. Journal of Autism Developmental Disorders 25(4), 355- 379.
Dunham , P. , & Dunham , F. (1992). Lexical development during middle infancy: A mutually
driven infant- caregiver process. Developmental Psychology, 28(3), 414-420.
Eheart, B. K (1982). Mother- child interactions with nonretarded and mentally retarded
preschoolers. AmericanJournal of Mental Deficiency, 87(1), 20- 25.
Greenspan , S. 1. , & Wieder , S. (1998). The child with special needs. Reading, MA , US: Addison-
Wesley.
Greenspan , , & Wieder , S. (1999). A functional developmental approach to autism spec-
S. 1.
trum disorders. Journal of the Association for Persons with Severe Handicaps 24 (3), 147- 161.
Jones , O. L. M. (1977). Mother- child communication with pre- linguistic Down s syndrome and
normal infants. In H. R Schaffer (Ed. Studies in mother- inpnt interaction. New York , NY
US: Academic Press.
Kasari , C. , Sigman , M. , Mundy, P. , & Yirmiya , N. (1988). Caregiver interactions with autistic
children. Journal of Abnormal Child Psychology, 16(1), 45-56.
Kaye , K (1977). Toward the origin of dialogue. In H. R Schaffer (Ed. Studies in mother- inpnt
interaction (pp. 89- 117). London , England: Academic Press.
Kaye , K (1982). The mental and social lif of babies: How parents create persons. Chicago , IL , US:
University of Chicago Press.
Kaye , K , & Fogel , A. (1980). The temporal structure of face- to- face communication between
mothers and infants. Developmental Psychology, 16(5), 454-464.
Koegel , L. K , Koegel , R L. , Harrower j. K , & Carter , C. M. (1999). Pivotal response inter-
vention I: Overview of approach. Journal of the Association for Persons with Severe
Handicaps 24(3), 174- 185.
Koegel , L. K , Koegel , R L. , Shoshan , Y , & McNerney, E. (1999). Pivotal response interven-
tion II: Preliminary long- term outcomes data. Journal of thlt Association for PItSOf!
with Sltllltlt Handicaps, 24(3), 186- 198.
Krug, D. A. , Arick , j. R , & Almond , P. j. (1980). Autism Screening Instrument ftr Education
Planning. Portland , OR , US: AISEP Educational Company.
), ),
Laski , K. E. , Charlop, M. H. , & Schreibman , L. (1988). Training parents to use the Natural
Language Paradigm to increase their autistic children s speech. Journal of Applied Behavior
Analysis 21(4), 391- 400.
Ie Couteur , A. , Rutter, M. , Lord , C. , & Rios , P. (1989). Autism Diagnostic Interview: A stan-
dardized investigator-based instrument. Journal of Autism Developmental Disorders, 19(3),
363-387.
Lord , C. , & Rhea , P. (1997). Language and communication in autism. In D. j. Cohen & F. R.
Volkmar (Eds. Handbook of autism. New York , NY , US: Wiley & Sons.
Lord , C. , Rutter , M. , & Le Couteur , A (1994). Autism Diagnostic Interview - Revised: A
revised version of a diagnostic interview for caregivers of individuals with possible per-
vasive developmental disorders. Journal of Autism Developmental Disorders, 24(5),
659- 685.
Loveland , K. A , & Landry, S. H. (1986). Joint attention and language in autism and develop-
mental language delay. Journal of Autism Developmental Disorders 16(3), 335- 349.
Markus , j., Mundy, P. , Morales , M. , Delgado , C. E. F. , & Yale , M. (2000). Individual differ-
ences in infant skils as predictors of child- caregiver joint attention and language. Social
Development 9 (3), 302- 315.
McGee , G. G. , Morrier , M. j., & Daly, T. (1999). An incidental teaching approach to early
intervention for toddlers with autism. Journal of the Association for Persons with Severe
Handicaps 24(3), 133- 146.
Mullen , E. M. (1995). Manual ftr the Mullen Scales of Early Learning. Circle Pines , MN , US:
American Guidance.
Mundy, P. , Hogan , A E. , & Doehring, P. (1996). A preliminary manual ftr the abridged Early
Social Communication Scale (ESCS). Unpublished manuscript.
Mundy, P. , Sigman , M. D. , Ungerer j., & Sherman , T. (1986). Defining the social deficits of
autism: The contribution of non-verbal communication measures. Journal of Child
Psychology Psychiatry Allied Disciplines 27(5), 657- 669.
Newson j. (1978). Dialogue and development. In A Lock (Ed. Action, gesture, and symbol: The
emergence of language. London , UK: Academic Press.
Ninio , A. , & Bruner , j. (1978). The achievement and antecedents of labelling. Journal of Child
Language 5(1), 1- 15.
Connell , R. j. (1969). St. Augustine s conftssions: The Odyssey of soul. Cambridge , MA , US:
Belknap Press of Harvard University Press.
Ratner , N. , & Bruner j. (1978). Games , social exchange and the acquisition oflanguage. Journal
of Child Language 5(3), 391- 401.
Reynell j. (1977). Reynell Developmental Language Scales. Windsor , England: NFER Publishing.
Robins , D. L. , Fein , D. , Barton , M. L. , & Green , j. A (2001). The Modified Checklist for
Autism in Toddlers: An initial study investigating the early detection of autism and per-
vasive developmental disorders. Journal of Autism and Developmental Disorders, 31(2),
131- 144.
Rogers , S. j., Ozonoff, S. , & Maslin- Cole , C. (1991). A comparative study of attachment behav-
ior in young children with autism or other psychiatric disorders. Journal of the American
Academy of Child Adolescent Psychiatry, 30(3), 483-488.
Rogers , S. j., Ozonoff, S. , & Maslin- Cole , C. (1993). Developmental aspects of attachment
behavior in young children with pervasive developmental disorders. Journal of the
American Academy of Child Adolescent Psychiatry, 32(6), 1274- 1282.
Schopler , E. , Reichler , R. j., & Renner , B. R. (1986). The Childhood Autism Rating Scale. New
York , NY , US: Irvington Publishers.
Schreibman , L. , Kaneko , W M. , & Koegel , R. L. (1991). Positive affect of parents of autistic
children: A comparison across two teaching techniques. Behavior Therapy, 22(4),
479- 490.
001- 018 10/28/03 12: 19 PM Page 17
177235-
Watson , L. R. (1998). Following the child' s lead: Mothers ' interactions with children with
autism. Journal of Autism Developmental Disorders 28(1), 51- 59.
Wiig, E. H. , Secord , W. , & Semel , E. (1992). Clinical Evaluation of Language Fundamentals-
Preschool. San Antonio , TX , US: The Psychological Corporation.
001- 018 10/28/03 12: 19 PM Page 18
177235-
Mailing Address:
Acknowledgements
Preparation of this article was supported by the Program Project Grant HD-
DCD35470 , Grant HD 17662 from the National Institute of Child Health and Human
Development and the National Institute of Deafness and Communication Disorders
Grant NS25243 from the National Institute of Neurological Diseases and Stroke , and
the M. I.N. D. Institute Research Program.
We wish to thank the children who allowed us to observe them growing up. We
extend our appreciation to their parents who committed so much time and effort to
our project.
We also wish to thank all the researchers who contributed to this project over the
years.
77235- 019- 026 10/16/03 4 : 26 Page 19
Abstract. Within the field of autism spectrum disorder, the attribute 'scientifcally proven ' is
most commonly seen in reference to the results of early behavioral treatment, and in particular
one style of early behavioral treatment. In this brief article, such claims are evaluated. Concerns
raised by other researchers about the methodology of the original Lovaas (7987) study are briefly
summarized. A particular concern that has been raised repeatedly is the lack of random assign-
ment of particiPants to treatment versus control group. A more recent study (Smith, Groen
Wynn, 2000), which included the necessary random assignment of particiPants to treatment
versus control group and assessed multiple outcome measures, is reviewed. The results of the
Smith et al. (2000) study with random assignment appear to be less dramatic than the results
from the original Lovaas (7987) study.
The attribute ' scientifically proven ' is common bait for consumers. Over 20 books
advertised on Amazon. com contain in their title the phrase scientifically proven.
These books include titles promising readers that they can reverse heart disease
(Ornish , 1996); gain physical fitness without exercise (Stamford , 1990); become an
effective coach (Smith & Small , 1996); cure age- related memory decline (Crook &
Adderly, 1998); and create world peace (Roth , 1994). Over 300 000 websites prom-
ise scientifically proven solutions to a myriad of challenges, ranging from the scien-
tifically proven way to stop cancer (www. stopping- cancer- naturally. org/) to the
scientifically proven " best way to lace your shoes " (www. techdirt.com/arti-
cles/20021204/1436200. shtml). There is even a U.S. political party (www. natural-
law. org) whose members claim that their platform is based solely on " scientifically
proven solutions to the nation s problems.
Of those 300 000 websites promising scientifically proven solutions , over 600
discuss scientifically proven ' solutions ' for autism. Frequent among these websites
about autism are the following claims: " FACT: There is a scientifically proven effec-
tive treatment principle for treating children with autism. This treatment is called
Applied Behavior Analysis (ABA). Using highly trained certified staff administering
30- 40 hours per week of intensive one- to- one treatment , the studies show that 47% of
the children reach normal functioning. They are indistinguishable from their peers.
(www. oregonparen tsunited. org/ articles/ effective _ autism treatment)
, "
Indeed , many agencies and individuals claim that only one style of early behavioral
intervention for autism is scientifically proven. For instance , BridgesABAtapes. com
(a company that sells audio tapes for ABA training) claims that " although parents of
autistic children are constantly bombarded by theories claiming to cure autism , only
one treatment is passing the test of time and research: Applied Behavior Analysis
(ABA)" (www. bridgesabatapes. com/autism. html). A student at Drury University
reporting about her summer internship claims that "ABA therapy is the only scien-
tifically proven and documented way of enabling preschoolers to enter grade school
indistinguishable from their peers " (www. drury. edu/multinl/story. cfm?ID=4397&
NLID=202). In his online comment about Maurice et al.' s (1996) book , a reader-
reviewer on the Barnes & Noble website claims: " Behavioral therapy, the discrete
trial method used by Dr. Lovaas , is the only scientifically proven treatment for
autism.
The sense of singularity among some individuals is so strong that the Clinical
Practice Guideline for Assessment and Intervention of Young Children with Autism/
Pervasive Developmental Disorder (1999), sponsored by the New York State
Department of Health , recommends that some interventions not even be included in
a child' s therapeutic program because those interventions might take time away from
an intervention that had been scientifically proven. Behavior Analysis , Inc omi-
nously warns that " diverting attention , even for a brief period of time , away from
treatment methods that have been scientifically proven to be effective is a disservice
and can have serious consequences " (www. behavior- analysis. org).
But what is the science behind these claims that one style of early behavioral
intervention for autism is ' scientifically proven?' Are there , as stated on the Surgeon
General's website thirty years of research demonstrating the efficacy of applied
behavioral methods in reducing inappropriate behavior and in increasing communi-
cation , learning, and appropriate social behavior " (www. surgeongeneral.gov/library/
mentalhealth/chapter3/sec6. html#autism)? This question can be answered by refer-
ring to the New York State Department of Health Guideline for Assessment and
Intervention of Young Children with Autism/Pervasive Developmental Disorder
(1999) because , in formulating their guideline , the authors conducted a thorough lit-
erature search. The authors found " 232 articles that reported using behavioral and
educational approaches in children with autism. ... These articles were systematically
screened and five articles (reporting four studies) were found that met the (authors
established criteria for adequate evidence about efficacy " (p. IV- 17).
Thus , of the 232 articles that the authors of the New York State Guideline found
in their exhaustive literature search ,only five articles met their own standards for
adequate evidence. And those five articles report only four studies. Those four stud-
ies are what can now be called the classic study by Lovaas (1987) published over fif-
teen years ago , a study by Birnbrauer and Leach (1993), a study by Smith and
colleagues (1997), and a study by Sheinkopf and Siegel (1998). Those four studies are
the only scientific proof that met the New York State Department of Health Clinical
Practice Guideline s criteria for adequate evidence.
,"
019- 026 10/16/03 4 : 26 Page 21
177235-
However , as even the New York State Guideline state: " None of the four studies
that met criteria for effcacy used random assignment of the children to groups (such
as to the group receiving intensive behavioral intervention versus the group receiv-
ing a comparison intervention)" (p. IV- 22). Random assignment is the practice of
assigning participants to conditions (e. , treatment versus control) such that each
participant has an equal chance of being assigned to each of the conditions. As any
scientist knows , random assignment is a core feature of scientific credibility in treat-
ment studies.
Unfortunately, however , even though Lovaas ' (1987) study did include an exper-
imental and a control group- indeed , there were two intended control groups-
assignment to either the experimental or the control group was not random. Rather
as the authors described , assignment to the experimental versus control group was
based on the therapists ' availability. Birnbrauer and Leach' s (1993) study also
included an experimental and a control group; unfortunately, as with the Lovaas
(1987) study, assignment to either the experimental or the control group was again
not random. In this case it was based on what the authors called " practical factors.
And neither Smith et aI's (1997) study nor Sheinkopf and Siegel's (1998) study used
a classic experimental design. Rather. both used retrospective data (i. , once the out-
comes were known , the authors looked back in time to see which treatment the sub-
jects had received; therefore , by definition , subjects were not assigned randomly to
those treatments). Thus , only the Lovaas (1987) and the Birnbrauer and Leach (1983)
studies qualify as true experimental designs , but disappointingly neither used ran-
dom assignment, which is a prerequisite for empirical interpretation.
Although the New York State Guideline authors suggest that " it has been argued
that the (non- random) method for group assignment probably did not bias the
results " (p. IV- 22), many scientists would draw the same conclusions as those drawn
in a recent article titled Separating fact from fiction in the etiology and treatment of
autism: A scientific review of the evidence " which was published in the Scientifc
Review of Mental Health Practice (Herbert , Sharp, & Gaudiano , 2002; see also Foxx
1993; Kazdin , 1993; Schopler , Short , & Mesibov , 1989). The Scientifc Review of
Mental Health Practice article suggests that the " methodological weaknesses of the
existing studies , however , severely limit the conclusions that can be drawn about
their efficacy. ... Of particular note is the fact that no study to date has utilized a true
experimental design , in which subjects were randomly assigned to treatment condi-
tions " (p. 37). Indeed , Herbert and Brandsma (2001) wrote , in an editorial titled
Applied Behavioral Analysis for childhood autism: Does the emperor have
clothes? " published in the Behavior Analyst Today: Most critically, the Lovaas study
was not a true experiment, as participants were not randomly assigned to (treatment
versus control) groups. The manner in which subjects were assigned to groups raises
serious questions about the possibility of selection bias , which are underscored by
pre- intervention differences between the experimental and control groups. These
methodological weaknesses limit the conclusions that can be drawn from this hall-
mark study (p. 47).
, p.
019- 026 10/16/03 4 : 26 Page 22
177235-
However , in 2000 , Smith , Groen , and Wynn published the first truly randomized
trial of intensive early intervention for children with pervasive developmental disor-
der. As the authors wrote: " To address criticisms of previous research and increase
methodological rigor , we conducted a fully randomized clinical trial with uniform
comprehensive assessment protocols for all participants " (p. 271). Smith et al.' s (2000)
experiment attempted to overcome other criticisms , as well. Despite claims that early
intervention based on the principles of ABA can produce " large , comprehensive
lasting and meaningful improvements in many important domains " (Green , 1996
38), the original Lovaas (1987) study included only two outcome measures: post-
treatment IQ scores and public school placement. Changes in IQ could reflect
increased compliance with testing rather than true changes in cognitive abilities , and
school placement could have more to do with parent advocacy and differential
school policies than with actual functional changes. Thus , Smith et al.' s (2000) first
randomized treatment study overcame these limitations by including assessment of
several important domains of functioning.
Smith et al.' s (2000) study involved 28 children , whose age at intake ranged from
24 months to 43 months and whose age at follow up ranged from 41 months to 117
months. Fifteen children were randomly assigned to the treatment group, and 13 chil-
dren were randomly assigned to the control group. The children randomly assigned
to the treatment group received Lovaas style intervention for an average of 25 hours
per week lasting a range of 18 months to 63 months. The children randomly assigned
to the comparison group received intervention as delivered by their parents. In other
words , the comparison group received parent- instructed treatment. At either intake
or at follow- up, Smith et al. measured these domains: Intelligence , Academic
Achievement, Language , Socioemotional Functioning, and Adaptive Functioning.
The researchers measured intelligence using either the Stanford- Binet
Intelligence Scales (Thorndike , Hagen , & Sattler , 1986) or the Bayley Scales ofInfant
Development- Mental Development (Bayley, 1969) and the Merril Palmer Scale of
Mental Tests (Stutsman , 1948). The researchers measured academic achievement via
Wechsler Individualized Achievement Test (WIAT). The WIAT was administered
only at follow up. The researchers measured language via the Reynell
Developmental Scales (Reynell , 1990), which have a scale for expressive language -
how well the child produces language- and receptive language- how well the child
comprehends language. The researchers measured socioemotional functioning via
the Achenbach Child Behavior Checklist (Achenbach , 1991). This checklist , which
was also administered only at follow up, was completed by both the child' s primary
caregiver and the child' s teacher. The checklist covers issues such as social with-
drawal , social problems , attention problems , and behavior problems such as aggres-
sion. The researchers measured " adaptive functioning " using the Vineland Adaptive
Behavior Scales (Sparrow , Balla , & Cicchetti , 1984). These scales are derived from an
interview with the primary caregiver. The study used three scales: Communication
Daily Living Skils , and Socialization. Adaptive functioning was assessed at both
intake and follow up.
019- 026 10/16/03 4 : 26 Page 23
177235-
As Smith et al. (2000) report: " Two of the 15 intensively treated children met the
criteria used by Lovaas (1987) and McEachin et al. (1993) for classifying children as
best outcome,' namely placement in regular classes without special services and
IQ/85). " Thus , in contrast to Lovaas ' (1987) reported 47% success rate according to
their two outcome measures of IQ and school placement for children treated with
Lovaas- style intervention , using the scientifically crucial random assignment Smith et
al. reported only a 13% success rate.
The other outcome measures were also substantially less dramatic. There were
statistically significant differences between the Lovaas- style treatment group and the
parent-based treatment group at follow up on both the Stanford- Binet and the Merril
Palmer. There was only a marginally statistically significant difference between the
Lovaas- style treatment group and the parent- based treatment group on the measure
of academic achievement. However , this statistical analysis might be somewhat com-
promised because nearly one- third of the control group was missing data on this test.
There were no statistically significant differences between the Lovaas-style treatment
group and the parent- based treatment group on either of the two language scales.
Although this is reported to be a significant difference in the paper , there was an
error in data analysis (and an erratum was subsequently published , Smith , 2001).
There were also no statistically significant differences between the Lovaas- style treat-
ment group and the parent-based treatment group in socioemotional functioning,
assessed by the Achenbach Child Behavior Checklist. And there were no statistically
significant differences between the Lovaas- style treatment group and the parent-
based treatment group in adaptive functioning as assessed by the Vineland Scales of
Communication , Daily Living Skils or Socialization or even as assessed by a com-
posite of those three scales. Indeed , even within the Lovaas- style treatment group
there were no gains in adaptive functioning from intake to follow up.
Smith et al. (2000) should be heartily applauded for undertaking the definitive
test of Lovaas- style intervention using the crucial ingredient of randomized assign-
ment. Such an experiment was far from easy to conduct; if it were , others would have
done it much earlier. Most strikingly, no other intervention has been submitted to
such empirical scrutiny. It is to the great credit of ABA proponents that they have
consistently sought to provide scientific evidence of the efficacy of their treatment.
However , given these data , namely that only one area of assessment showed a sta-
tistically significant difference due to treatment , and that only 13% met the criteria of
success outlined by Lovaas (1987), it is perhaps appropriate to agree with the follow-
ing admonition from the Scientifc Review of Mental Health Practice: Given the current
state of the science , claims of ' cure' and ' recovery ' from autism produced by ABA
are misleading and irresponsible " (p. 37).
A recent New York Times article about intervention for autism (Tarkan , October
2002) began by stating that " no one has found a cure for autism , the neurological
disorder that leads to lifelong impairments in a child' s ability to speak , respond to
others , share affection and learn. But there is a growing consensus that intensive early
intervention is both effective and essential- the sooner after diagnosis , the better.
Early intervention , which involves many hours of therapy with one or more special-
),
ists (NB: the article later states that there are several different types of therapy) does
not help every autistic child to the same degree... and for reasons that are unclear
it does not help some children at all. But for those who are helped , their parents say,
the changes are miraculous.
It behooves all of us to find the avenues that wil lead to what every parent would
call miraculous. In route to finding those avenues , we should most likely exercise
caution in claiming that one style of intervention has been scientifically proven.
References
Achenbach , T. M. (1991). Integrative guide for the 7997 Child Behavior Checklist/4- , YSR, and
Teacher Report Form profiles. Burlington: University of Vermont , Department of
Psychiatry.
Bayley, N. (1969).Bayley Scales of Infant Development. New York: Psychological Corporation.
Birnbrauer j., & Leach , D. (1993). The Murdoch early intervention program after 2 years.
Behavior Change 63- 74.
Crook , T. , & Adderly, B. D. (1998). The memory cure: The safe, scientifcally proven breakthrough that
can slow, halt, New York: Pocket Star Books.
or even reverse age- related memory.
autism: A manual for parents and professionals. Austin , TX: PRO- ED.
McEachin j., Smith , T. , & Lovaas , O. 1. (1993). Long- term outcome for children with autism
who received early intensive behavioral treatment. AmericanJournal of Mental Retardation
359-372.
New York State Department of Health Early Intervention Program. (1999). Clinical practice
guideline: The guideline technical report. Autism/Pervasive Developmental Disorders.
Assessment and intervention for young children (age 0- 3 years). New York , NY.
Ornish , D. (1996). Dr. Dean Ornish's program for reversing heart disease: The only system scientifcally
proven to reverse heart disease without drugs or surgery. New York: Ivy Books.
Reynell , j. K. (1990). Reynell Developmental Language Scales. Los Angeles , CA: Western
Psychological Corporation.
Roth , R (1994). TM- Transcendental Meditation: A new introduction to Maharishi' s easy, efctive and
scientifcally proven technique ftr promoting better health, unfolding your creative potential, and
creating peace in the world. New York: Fine.
Schopler , E. Short, A. , & Mesibov , G. (1989). Relation of behavioral tratement to " normal
functioning; " Comment on Lovaas. Journal of Consulting and Clinical Psychology, 47
162- 164.
019- 026 10/16/03 4 : 26 Page 25
177235-
Sheinkopf, S., & Siegel , B. (1998). Home-based behavioral treatment of young children with
autism. Journal of Autism and Developmental Disorders , 15- 23.
Smith , R. E. , & Small , F. L. (1996). fVy to go, coach: A scientifcally proven approach to coaching effc-
tiveness. Portola , CA: Warde Publishers.
Smith , T. , Eikeseth , S. Klevstrand , M. , & Lovaas , O. 1. (1997). Intensive behavioral treatment
for preschoolers with severe mental retardation and pervasive developmental disrder.
AmericanJournalof Mental Retardation 102 , 238-249.
Smith , T. , Groen , A. D. , Wynn j. W. (2000). Randomized trial of intensive early intervention
for children with pervasive developmental disorder. American Journal of Mental
Retardation 105 269- 85.
Smith , T. , Groen , A. D. , Wynn j. W. (2001). Erratum. AmericanJournal of Mental Retardation
106 , 208.
Sparro , S. , Balla, D. A. , & Cicchetti ,
D. V. (1984). The Vineland Adaptive Behavior Scales. Circle
River , MN: American Guidance Service.
Stamford , B. (1990). Fitness without exercise: The scientifcally proven strategy for achieving maximum
health with minimum effort. New York , NY: Warner Books.
Stutsman , R. (1948). Guide for administering the Merrill-Palmer Scale of Mental Tests. New York:
Harcourt, Brace & World.
Tarkan , L. (2002). Autism therapy is called effective , but rare. New York Times.
Thorndike , R. L. , Hagen , E. P. , & Sattler j. M. (1986). Stanford-Binet Intelligence Scale (4th ed.
Chicago: Riverside.
Mailing Address:
University of Wisconsin-Madison
7202 WJohnson Street
Madison, WI 53706- 7677
(608) 262- 6989 ffx (608) 262- 4029)
http://psych. wisc. edu/lang/MGcover. html
019- 026 10/16/03 4:26 PM Page 26
177235-
77235- 027- 036 10/17/03 9:11 AM Page 27
ing when speech stimuli were acoustically or lexically degraded. Concurrent tests of auditory
working memory, tone discrimination, and word recognition were normal. These results suggest
subtle impairments in the fUndamental speech perception abilities of children with AS
INIODUCTION
Asperger syndrome (AS) is a neurodevelopmental disorder characterized by
impaired social behavior and focused interests , in the context of seemingly normal
language and cognitive development (American Psychiatric Association , 1994).
Although social aspects of language are impaired in AS, as in most autism spectrum
disorders (Landa , 2000), other language functions appear normal. The reported
absence of language delays in AS and the observation that language abilities in these
often highly verbal individuals are typically better than in other autism spectrum dis-
orders support this view (Baltaxe & Simmons , 1985; Tager- Flusberg, 1981; Wing,
1981).
Recent studies , however , have revealed subtle language abnormalities in AS.
Using fine- grained analyses , Shriberg and colleagues (2001) identified articulation
"Portions of this paper were presented at the Interdisciplinar Council on Developmental and Learning
Disorders Sixth Annual International Conference on Autism and Disorders of Relating and Communicating in
McLean , Virginia on November 9, 2002.
027- 036 10/17/03 9:11 AM Page 28
177235-
DANA BOATMAN
distortions in 33% of their AS subjects. Higher- level receptive and expressive lan-
guage abnormalities have also been observed (Howlin , 2003; Koning & Magill-
Evans , 2001). To date , little is known about basic speech perception abilities in AS.
Deficiencies in early processing of spoken speech , including the ability to decode
consonant and vowel speech sounds (phonemes), can impact language functions , as
observed in children with developmental language disorders (Tallal , Stark , & Mellits
1985). Moreover , it has been shown that children with autism fail to show the same
preference for speech sounds as normally developing children (Klin , 1991). In light
of these reports , investigation of phoneme processing in AS appears warranted. In a
preliminary attempt to address this issue , we tested multiple aspects of phoneme pro-
cessing in six children with AS.
METIODS
Parcipants
Six children (age 9- 14; mean age 10;6 years), five boys and one girl participated
(Table 1). All met DSM- IV criteria for AS , as determined by a pediatric neurologist
(American Psychiatric Association , 1994). Diagnostic criteria included severe social-
behavioral impairments , focused interests , and normal language and cognitive devel-
opment. Four were right handed and two were left handed. Normal language
development was defined clinically as the absence of language delays or regression
with production of 2- 3 word phrases by age 3 years (American Psychiatric
Association , 1994). On neuropsychological testing, all participants obtained full- scale
IQ scores 93 (mean = 118 , SD :t18. 7) on the Wechsler Intelligence Scale (WISC-
III R , Wechsler , 1991). All had normal neurological examinations , negative tests for
Fragile X , no co- morbid neurological disorders (e. , seizures), no history of chronic
otitis media , and no history of hearing or speech disorders. Because the presence of
a hearing loss can interfere with detection of acoustic cues used to process phonemes
(Schuknecht, 1974), only AS subjects with normal hearing were included. Subjects
with a history of chronic otitis media , defined as more than two ear infections by age
6 years , were also excluded because of an association with hearing loss. Two partici-
Procedures
All testing was performed in a sound- treated audiometric booth using a two- chan-
nel audiometer (GSI- 30) and an acoustic immittance system (Madsen Zodiac 901).
Testing was completed in one session , with the number of breaks determined by indi-
vidual needs. Parental consent was obtained for all participants in accordance with
our institution s IRB requirements.
Hearng evaluations
All participants underwent routine hearing evaluations to confirm normal hear-
ing. Otoscopic exams were performed to rule out excessive cerumen or foreign bod-
ies in the ear canal that might interfere with test results. Pure tone air and bone
conduction thresholds were established behaviorally for each ear at seven octave and
inter- octave frequencies from 0. 25- 8 kHz. Participants pressed a button in response
to sound presentation. Results were interpreted based on published clinical norms of
s: 20 dB HL as the range of normal hearing thresholds (ASHA , 1978). In light of con-
cerns that examiner bias and subject factors can affect testing of individuals with
autism spectrum disorders (Gravel et al. , 1994), Distortion Product Otoacoustic
Emissions (DPOAEs) were performed to provide an objective measure of the fre-
quency- specific responses of outer hair cells (Kemp, 1978). A handheld screener
(ERO SCAN , Etymotic Research) was used with an eartip probe that fit into the ear
canal with a seal. DPOAEs were elicited for each ear at six frequencies (1.5 , 2 , 3 , 4
5 and 6 kHz). Emissions were measured in the 2fl- f2 frequency band (fl intensity =
65 dB SPL , f2 intensity = 55 dB SPL), using a pass criterion of 5 dB or greater above
the average noise level in adjacent frequency bands. Speech reception thresholds
027- 036 10/17/03 9:11 AM Page 30
177235-
DANA BOATMAN
were determined with spondee words , and word recognition was tested in quiet using
CID W22 word lists (40 dB SL , 25 words per ear). Tympanometry (226 Hz probe),
performed to assess middle ear function , yielded tympanograms classified as normal
(Type A , :t150 daPa) or abnormal (Types B , C). Acoustic middle ear reflexes were
tested (2000 Hz , contralateral stimulation) and determined to be normal if they
occurred at ,s 90 dB SPL.
Auditory brainstem response (ABR) testing was also performed in light of previ-
ous reports of brain stem auditory dysfunction in children with autism spectrum dis-
orders (Ornitz , 1974; Tanguay et al. , 1982; Rosenhall et al. , 2003). Two participants
had undergone ABR testing previously, with normal results. For the four remaining
participants , ABRs were elicited with rarefaction clicks presented (21/sec) monau-
rally at 75 dB SPL (?: 55 dB SL) with 45 dB of masking noise in the contralateral ear.
Four silver disk electrodes were used to record from vertex (Cz) to the mastoid (AI
A2) ipsilateral to the stimulated ear , with a low forehead electrode (Fp) as ground.
Electrode impedances were maintained below 5 kOhm. The EEG filter was set to
bandpass 100- 3000 Hz. A total of 2000 trials were recorded for each ear using a
Nicolet Spirit Evoked Response System. For each participant , waveforms were aver-
aged using a post- stimulus analysis time of 15 msec. Peak latencies for waves I-V were
identified for each ear. Absolute and inter- peak latencies were compared to clinical
norms for subjects ' age ranges.
Phoneme processing
Three tests of auditory phoneme processing were administered. All used single
words or syllables produced by male native speakers of English. Test stimuli were
presented from compact disc through insert earphones (Eartone , 3A), at 50 dB HL
following 2- 4 practice items.
Phoneme blending. The standardized Phoneme Synthesis Test was used to assess
participants ' ability to blend individual phonemes into words (Katz & Fletcher , 1998).
Twenty- five phoneme strings , each containing 3-4 individual phonemes , were pre-
sented binaurally. Participants said the words they derived by blending each
phoneme string. For each participant , the total number of correctly synthesized
words was compared to published age norms.
Phoneme closure. The standardized Filtered Word subtest of the SCAN- CIA test
was administered to evaluate participants ' ability to compensate for loss of high- fre-
quency phonological cues resulting from low- pass acoustic filtering (Keith , 1994
2000). The children s version (SCAN- C) was administered to the four participants ,s
11 years of age , and the SCAN- A was administered to the two older children.
Participants repeated forty monosyllabic low- pass filtered words (1.0 kHz cut off, 32
dB/octave roll- off) presented to each ear. Each participant's score was compared to
published age norms.
Phoneme discrimination. A binary, forced- choice , same- different paradigm in which
twenty- five digitized (44 kHz , 16 bit sampling) consonant- vowel syllable pairs (e.
027- 036 10/17/03 9:11 AM Page 31
177235-
pa- , di- di) were administered binaurally. This test has been used with other popu-
lations (Boatman , Lesser , & Gordon , 1995; Boatman , Vining, Freeman , & Carson
2003). Eighteen syllable pairs differed in initial consonant. The remaining seven
pairs were identical. For this preliminary investigation , only consonant contrasts
were used because of their known association with phoneme processing (Boatman et
al. , 1995; Studdert- Kennedy & Shankweiler , 1970). Participants responded by point-
ing to one of two colored shapes representing ' same ' or ' different'. Because the
phoneme blending and closure tests both used verbal responses , we chose a non-
verbal response for the phoneme discrimination test to control for possible speech-
articulation effects. To provide a comparative basis for evaluating participants ' per-
formances , 6 control subjects were selected from a pool of normal subjects recruited
previously to establish test norms. They were matched to our AS subjects by age
gender , and handedness. All of the normal controls had 1) negative histories for neu-
rological, speech- language , or hearing disorders; 2) confirmed normal hearing bilat-
erally; and 3) normal scores on the SCAN- CIA and Phoneme Synthesis tests.
Other testing
Tone discrimination. A control test consisting of twenty pairs of frequency- modu-
lated (FM) tones (starting frequencies = 0. , 0. 9 Hz , slope = 0 , 0.4 , 0. 8 kHz) was
administered from compact disc using the same paradigm as the phoneme discrimi-
nation test. FM tones were selected over steady- state tones because of their compa-
rability to the consonant- vowel transitions of the phoneme discrimination stimuli.
Seven tone pairs were contrasted by direction of the frequency change (e. , rising
versus falling), seven by slope of the frequency change (e. , 0.4 Hz versus 0. 8 kHz),
and six were identical. One participant (Participant 6) was evaluated before the tone
discrimination test was added to the experimental protocol. Individual scores were
compared with results from the same normal controls tested on the phoneme dis-
crimination test.
Digit Span test. To assess auditory (verbal) working memory, participants repeated
(forward , backward) digit sequences of increasing length that were administered live-
voice from the Wechsler intelligence scales (Wechsler , 1991).
RESULTS
Participants were fully compliant and completed all testing. Test results are sum-
marized in Table 2. All participants demonstrated normal pure tone air conduction
(-s 20 dB HL , 0. 25- 8 kHz) and bone conduction (-s 20 dB HL , 0. 4 kHz) thresholds
bilaterally. All met DPOAE pass criteria for both ears, confirming the normal behav-
ioral threshold findings. For all participants , speech reception thresholds were within
10 dB of pure tone averages (0. 2 kHz), suggesting good response reliability, and
word recognition in quiet was within the normal range (?: 92%). Type A tym-
(p
DANA BOATMAN
Other Tests
FM Tone Discrimination 1.0 1.0 85- 1.0
Digit Span Test
"Phoneme blending as measured by the Phoneme Synthesis Test (Katz & Fletcher , 1998)
bLower limit of normal for age 9 = . 72 on the Phoneme Synthesis Test (Katz & Fletcher , 1998)
cLower limit of normal for age 10 = . 84 on the Phoneme Synthesis Test (Katz & Fletcher , 1998)
dPhoneme closure measured by the Filtered word subtest of the
SCAN- CIA (Keith , 1996
2000)
eLower normal limit for age 9 = . 75 on the Filtered Word Subtest of the SCAN- C (Keith , 2000)
fLower normal limit for age 10 = . 77 on the Filtered Word Subtest of the SCAN- C (Keith , 2000)
gTest results given as scaled scores (Mean 10 , SD :t 3)
NA = not administered
panograms were obtained bilaterally in all cases , suggesting normal middle ear func-
tion , and acoustic reflexes were present at normal levels. ABR recordings yielded
reproducible waveform morphologies. Based on clinical norms , absolute and inter-
peak latencies for waves I -V were determined to be within the normal range at stim-
ulus presentation levels of 75 dB , with no significant inter- aural differences in wave
V latencies.
All participants performed below normal limits for their age on tests of phoneme
processing. On a test of phoneme blending (Phoneme Synthesis Test), five partici-
pants performed below the normal range (92- 100%), with an average score of 64%
(range 24- 88% , SD :t24.l), and one participant (Participant 1) performed at the
lower limit of normal for his age. AS participants demonstrated impaired phoneme
closure abilities , as measured by the Filtered Word test. Compared with normal sub-
jects who obtained a mean score of 88% (range = 80- 95% , SD :t7), the average score
for AS participants was 60% (range = 48- 76% , SD :t 12), with five performing three
or more standard deviations below the normal range. Response errors on the
phoneme blending and closure tests consisted largely of consonant substitutions
andlor deletions (e. , pie;: buy, hit;: hop, need;: knee). On the phoneme dis-
crimination test , AS participants ' scores ranged from 76- 96% (mean 85% , SD :t 7.84),
while those of the normal controls ' ranged from 92- 100% (mean 98% , SD :t 3. 34).
Logistic regression was used to compare the two groups , allowing over dispersion in
the scale parameter to account for correlation within subjects (McCullagh & NeIder
1989). The AS participants performed significantly worse on the phoneme discrimi-
nation test than the normal controls = 0. 0002 , logistic regression).
(p
On the FM tone discrimination test, AS participants ' scores (mean 96% , range
90- 100% , SD :t1.09) did not differ significantly = 0. , logistic regression) from
those of the normal controls (mean 93% , range 85- 100% , SD :t1.04). There was con-
siderable overlap in the range of scores for both groups , with the AS participants per-
forming slightly higher overall. On the Digit Span test of auditory working memory,
five AS participants obtained scaled scores that were normal (10 :t 3), or better. One
participant (Participant 2) scored just below the lower limit of normal (scaled score = 6).
DISCUSSION
This study provides preliminary evidence that phoneme processing is impaired
in children with AS. The perceptual deficiencies observed in our six AS participants
were relatively subtle , emerging only when the speech stimuli were degraded acousti-
cally (phoneme closure) or lexically (phoneme blending, phoneme discrimination).
The subtle nature of these phoneme processing difficulties may explain why they
have previously gone undetected in AS. Behavioral evidence of phoneme processing
difficulties coupled with normal hearing and ABR results , suggests a dysfunction
higher in the auditory system. Lesion studies have associated phoneme processing
disorders with dysfunction of the left temporal lobe (Boatman et al. , 1995; Luria
1976; Miceli , Caltagirone , Gainotti , & Payer- Rigo , 1978). Neuroimaging and electro-
physiology studies have also identified left temporal lobe abnormalities in individu-
als with autism spectrum disorders (Bruneau , Roux , Adrien , & Barthelemy, 1999;
Jones & Kerwin , 1990; Zilbovicius , et al. , 2000). The concurrent sparing of fre-
quency- modulated (FM) tone discrimination further suggests that their perceptual
difficulties did not stem from a more general auditory dysfunction or from difficulty
comprehending task instructions. Differences in the acoustic complexity of the
speech versus tone stimuli , however , preclude any definitive claims about the speech-
specific nature of our participants ' perceptual difficulties. Although all were impaired
on measures of phoneme processing, there was considerable variability in their per-
formance. Additional studies with larger numbers of AS participants as well as nor-
mal controls wil help to determine the range of variability and the extent to which
there may be overlap with that of normal subjects.
Participants ' errors on the phoneme blending and closure tests consisted largely
of phoneme substitutions and/or deletions. Because these tests relied on verbal
responses , the possibility of confounding effects from articulation warrants consider-
ation. While previous speech- language evaluations revealed no evidence of overt
articulation disorders , we cannot rule out more subtle articulation difficulties in the
absence of detailed acoustic analyses , as recently suggested (Shriberg et al. , 2001).
Participants ' poor phoneme processing scores cannot be attributed entirely to speech
production difficulties , however , since they also performed poorly on the phoneme
discrimination test that did not involve verbal responses. It may be argued that an
impairment of auditory working memory would account for participants ' poor per-
formances on phoneme tests that used longer stimuli , including the phoneme blend-
027- 036 10/17/03 9:11 AM Page 34
177235-
DANA BOATMAN
ing and discrimination tests. However , all but one participant demonstrated normal
auditory working memory abilities on digit span testing, and all performed normally
on the tone discrimination test , which used the same presentation format as the
phoneme discrimination test.
Alternatively, an impairment of attention might account for the pattern of per-
ceptual deficits observed especially since the phoneme processing difficulties
emerged only under adverse listening conditions that also impose increased attention
demands. Although only one participant had been diagnosed with an attention dis-
order , we cannot rule- out the possibility of a mild , previously unidentified attention
disorder in the others. Because all perceptual tests administered were behavioral and
therefore , involved attention , we cannot rule- out potential effects of attention on par-
ticipants ' performances. Electrophysiological recordings offer one means of evaluat-
ing auditory processing independent of attention. A recent electrophysiology study
of speech sound processing in children with high- functioning autism found evidence
of a selective impairment in attention orienting to vowel sounds , with otherwise nor-
mal sensory processing of vowels and acoustically matched complex sounds
(Ceponiene et al. , 2003). This suggests that the impairment in speech sound percep-
tion occurs after early sensory processing and reflects deficiencies in attention.
Because children with very poor verbal language abilities were included , we cannot
make direct comparisons with our participants , who had normal verbal language
abilities. However , given the reportedly high rate of attention difficulties in AS
(Schatz , Weimer , & Trauner , 2002), this issue clearly warrants further investigation.
Our participants ' history of normal language development is surprising given
their observed phoneme processing difficulties. One possibility is that their phoneme
processing difficulties were sufficiently subtle as to not impede language develop-
ment. Alternatively, the predominantly single- word tests used to evaluate their recep-
tive and expressive language abilities may have lacked sufficient sensitivity, as
recently suggested (Konig & Magill- Evans , 2001). It is also possible that early delays
in language development resulting from perceptual deficiencies may have gone
undetected because formal testing is usually not performed before age 3 , by which
time these children may have begun to catch up with their peers (Howlin , 2003;
Landa , 2000). Further investigation is needed to identify AS children who would
benefit from early remediation of their phoneme processing and , in turn , other lan-
guage skils (Lindamood , Bell , & Lindamood , 1997; Merzenich , Jenkins , Johnston
Schreiner , Miler , & Tallal , 1996).
In summary, all six normal- hearing AS participants demonstrated impaired
phoneme processing abilities , as measured by tests of phoneme blending, closure
and discrimination. Their perceptual difficulties are not readily attributed to disor-
ders of the peripheral or brainstem auditory system , articulation , working memory,
or general auditory dysfunction. The potential contribution of attention difficulties to
the perceptual impairments observed remains to be determined. Future studies wil
help to clarify the exact nature of the phoneme processing difficulties observed , the
extent to which they may be modality specific and/or influenced by attention , and
their potential effects on language and cognitive function in AS.
), ),
REFERENCES
American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders
(4th ed. ). Washington , DC: American Psychiatric Association.
American Speech , Language , Hearing Association (ASHA). (1978). Guidelines for manual
pure- tone audiometry. Journal of the American Speech, Language and Hearing Association, 20
297- 301.
Baltaxe , C. , & Simmons j. (1985). Prosodic development in normal and autistic children. In
E. Schopler & G. Mesibov (Eds. Communication Problems in Autism (pp. 95- 125). New
York: Plenum Press.
Boatman , D. , Lesser , R , Gordon , B. (1995). Auditory speech processing in the left temporal
lobe: an electrical interference study. Brain and Language , 269- 90.
Boatman , D. , Vining, E. , Freeman , j., & Carson , B. (2003). Auditory processing studied
prospectively in two hemidecorticectomy patients. Journal of Child Neurology, 18(3),
228-232.
Bruneau , N. , Roux , S. , Adrien j., Barthelemy, C. (1999). Auditory associative cortex dysfunc-
tion in children with autism: evidence from late auditory evoked potentials (Nl wave-
complex). Clinical Neurophysiology, 110 , 1927- 1934.
Ceponiene , R , Lepisto , T. , Shestakova , A. , Vanhala , R , Alku , P. , Naatanen , R , & Yaguchi , K.
(2003). Speech- sound- selective auditory impairment in children with autism: they can
perceive but do not attend. Proceedings of the National Academy of Science, 100(9),
5567-5572.
Dunn , L. , & Dunn , M. (1981). Peabody Picture Vocabulary Test- Revised manual. American
Guidance Service: Circle Pines , MN.
Gardner , M. F. (1979). Expressive One- J;rd Picture Vocabulary Test manual. Novato , Academic
Therapy Publications.
Goldman , R , & Fristoe , M. (1969). Goldman- Fristoe Test of Articulation. Circle Pines , MN:
American Guidance Service.
Gravel j. (1994). Auditory integration training: placing the burden of proof. AmericanJournal
of Speech- Language Pathology, 3(2), 25- 29.
Howlin , P. (2003). Outcome in high- functioning adults with autism with and without early lan-
guage delays: implications for the differentiation between autism and Asperger
Syndrome. Journal of Autism and Developmental Disorders 33(1), 3- 13.
Jones , P., & Kerwin , R (1990). Left temporal lobe damage in Asperger s syndrome. British
Journal of Psychiatry, 156 , 570- 572.
Katz j., & Fletcher , C. (1998) Phonemic Synthesis Test. Vancouver , WA: Precision Acoustics.
Keith , R (1994). SCAN-A: A Testftr Auditory Processing Disorders in Adolescents and Adults. San
Antonio , TX: Psychological Corporation.
Keith ,R (2000). SCAN- C: A Test for Auditory Processing Disorders in Children: Revised. San
Antonio , TX: Psychological Corporation.
Kemp, D. (1978). Stimulated acoustic emissions from within the human auditory system.
Journal of the Acoustical Society of America , 1386- 1391.
Klin , A. (1991). Young autistic children s listening preferences in regard to speech: a possible
characterization of the symptom of social withdrawal. Journal of Autism and Developmental
Disorders 21(1), 29- 42.
Konig, C. & Magil- Evans , j. (2001). Social and language skils in adolescent boys with
Asperger syndrome. Autism 5(1), 23- 36.
Landa ,
R (2000). Social language use in Asperger Syndrome and high- functioning autism. In
Klin , A. , Volkmar , F. , & Sparrow , S. (Eds. Asperger Syndrome (pp. 125- 155). New York:
The Guilford Press.
Lindamood , P. , Bell , N. , & Lindamood , P. (1997). Sensory- cognitive factors in the controversy
over reading instruction. The Journal of Developmental and Learning Disorders 1 (1),
143- 182.
027- 036 10/17/03 9:11 AM Page 36
177235-
DANA BOATMAN
Mailing Address:
ACKNOWLEDGMENTS
This study was supported by the National Alliance for Autism Research. Special
thanks to Dr. D. Miglioretti for the statistical analyses , and to Drs. B. Gordon , A.
Zimmerman , and S. Reich for helpful discussion.
77235- 037- 056 10/28/03 12 : 24 Page 37
Keywords: Executive functioning, Autism , Inhibition , Dimensional change card sort task
Abstract. A group of 8-year- old children with autism was compared to a group of tYPically-
develoPing 4-year- olds (approximately matched for verbal mental age) and to a group of tyPi-
cally-develoPing 8-year- olds (matched for chronological age) on a computerized version of the
dimensional change card sort task. This task requires children to classif a set of stimuli first by
one dimension (pre- switch phase) and then to reclassif the same stimuli by another contrast-
ing dimension (post-switch phase). There were four conditions that difred in the nature of the
rule used to sort the stimuli and the processing demands required by the post-switch phase.
Children with autism were both quantitatively and qualitatively difrent from the other chil-
dren in the post- switch phase. They obtained the lowest scores across the four conditions and
were the only group to experience difculty in the simplest condition in which only a single per-
ceptual feature was used to sort the stimuli. Nonetheless, the children with autism solved the
most difcult condition based on a conceptual classifcation better than the tyPically- develoPing
year- olds, once vocabulary level was controlled. The results are discussed in terms of the role
of inhibition in the executive fUnctioning of children with autism.
For children with autism , cognitive deficits are only one aspect of a complex set of
symptoms , including ritualism , behavioural rigidity, perseveration , social difficulties
narrow focus , impulsivity, and communication impairments (American Psychiatric
Association , 1994). The cause of the cognitive deficits in particular is a matter of con-
troversy. Planning and regulation (e. , Adrien et al. , 1995; Hughes , 1996), metarep-
resentation (e. , Baron- Cohen , 1994; Baron- Cohen , Leslie , & Frith , 1985), inhibitory
control and selective attention (e. , Hughes & Russell , 1993; Ozonoff & Strayer
1997), working memory (e. , Boucher , 1981; Boucher & Lewis , 1989), and weak cen-
tral coherence (e. , Frith , 1989; Frith & Happe , 1994), have all been considered as
the locus of the cognitive deficit implicated in the pathogenesis of autism.
Although children with autism present a common, identifiable symptom profie (American Psychiatric
Association, 1994), the course , severity, aetiology, and expression of autism is heterogeneous (e. , Pelios
& Lund , 2001).
037- 056 10/16/03 5 : 09 Page 38
177235-
Two primary hypotheses have been proposed for the cognitive deficits of autism:
the theory of mind (TOM) or " mindblindness " hypothesis , and the executive dys-
function hypothesis. According to the TOM hypothesis , autism is caused by impair-
ments in intuiting and representing mental states on both an intrapersonal and an
interpersonal level (Baron- Cohen , 1989 , 1995; Perner , Frith , Leslie , & Leekam
1989). Advocates of this view contend that TOM deficits render the mutual under-
standings , shared feelings , surmised intentions , and jointly held beliefs characteristic
of everyday social relationships and interactions inaccessible to children with autism
(Baron- Cohen , 1989). This hypothesis was proposed by Baron- Cohen , Leslie , and
Frith (1985) when they discovered that many children with autism had diffculty in
passing the Sally- Anne false belief task (Wimmer & Perner , 1983), a classic measure
of TOM. Since then , extensive evidence has accumulated supporting this interpreta-
tion (Grant, Grayson , & Boucher , 2001; Happe , 1995; Or & Yirmiya , 1993; Yirmiya
Erel , Shaked , & Solomonica- Levi , 1998).
There is a substantial literature supporting this view , but several issues challenge
the TOM hypothesis. Foremost among these are concerns about reliability and valid-
ity. Regarding reliability, not all children with autism are characterized by TOM
deficits. For performance on first- order TOM tasks , Dahlgren and Trilingsgaard
(1996) report a 10% failure rate among 20 children with autism (age range = 6. 3 to
15. 5);Prior , Dahlstrom , and Squires (1990) report a 40% failure rate among 20 indi-
viduals with autism (age range = 5. 0 to 25. 0); and , Reed and Peterson (1990) report
an 85% failure rate among 13 participants (age range = 4. 3 to 29.1). Therefore , some
children with autism maintain at least some TOM abilities , undermining its causal
role in the disorder. Regarding validity, children with autism demonstrate proficien-
cies in daily living that would presumably be impossible with their alleged subnor-
mal TOM abilities (Frith , Happe , & Siddons , 1994; Happe , 1994). Some researchers
(e. , Happe , 1995; Hobson , 1991) contend that the poor performance of these chil-
dren on TOM tasks is (in part) an artefact of the structure of those tasks that chal-
lenges other difficulties that children with autism experience. Even children who pass
these tasks may do so by using compensatory strategies that enable them to mimic
appropriate responses without fully appreciating the mental states of others (Frith
Morton , & Leslie , 1991; Happe , 1993 , 1995).
The relation between TOM and autism may be mediated by language ability. A
study by Peterson and Siegal (1995) revealed that 17 of 26 patients with prelingual
deafness (age range = 8 to 23), all with normal intelligence , failed a simple TOM task.
Using more precise linguistic measures , de Viliers (1999) has identified the syntactic
ability to use that- complement clauses as the single best predictor of performance in
TOM tasks. She speculates that this linguistic achievement might be one factor in the
failure of children with autism to solve these tasks. Children with autism frequently
suffer from poorly developed language competence , and several studies have
demonstrated a positive relation between verbal mental age and performance on
TOM tasks (Eisenmajer & Prior , 1991; Leekam & Perner , 1991).
An alternative to the TOM hypothesis is that executive dysfunction is the pri-
mary explanation for the cognitive deficits typical of autism (Bryson , Landry, &
,&
037- 056 10/16/03 5 : 09 Page 39
177235-
Wainwright, 1997; Pennington & Ozonoff, 1996). ' Beginning with the seminal inves-
tigation of Hermelin and O' Connor (1970) and subsequently replicated in numerous
studies , it has been shown that individuals with autism consistently perseverate on a
well- known measure of executive dysfunction: the Wisconsin Card- Sorting Task
(WCST) (Bennetto , Pennington , & Rogers , 1996; Ozonoff, 1995; Ozonoff & McEvoy,
1994; Ozonoff et al. , 1991; Prior & Hoffman , 1990; Rumsey, 1985; Rumsey &
Hamburger , 1988; Shu , Lung, Tien , & Chen , 2001; Szatmari , Tuff, Finlayson
Bartolucci , 1990). Zelazo Jacques , Burack , and Frye (2002) extended these results by
showing that severely impaired children with autism perseverated on the dimen-
sional change card sort task , similar to the WCST , but that there was no relation
between their performance on the card sort task and TOM tasks. Perseveration on
these tasks is regarded as a failure to inhibit a well- rehearsed , prepotent response and
a tendency " to persist with a specific strategy, despite corrective feedback" (Ozonoff
1995; Shu et al. , 2001 , p. 170).
Executive dysfunction comprises impairments in goal- directed , future- oriented
mental operations that are mediated by the frontal lobes and involve planning, atten-
tion , inhibition , cognitive flexibility, working memory, and organized searching
(McEvoy, Rogers , & Pennington , 1993; Ozonoff, 1995; Pennington et al. , 1997).
Using criteria such as these , Ozonoff andJensen (1999) point out that the majority of
studies on children with autism have uncovered evidence of executive dysfunction.
In a wide- ranging battery of tasks administered to children with autism , attention-
deficit/hyperactivity disorder (ADHD), dyslexia , and mild mental retardation
Ozonoff et al. (1991) found that only the executive dysfunction measures (i. , the
WCST and the Tower of Hanoi) discriminated among groups. Although first- order
TOM tasks correctly classified 65% of the 23 children with autism (age range = 8.
to 20.0), executive dysfunction measures correctly classified 80% of the children with
autism.
Recent research has examined the cognitive architecture underlying the execu-
tive dysfunction hypothesis in an effort to determine the exact nature of the impair-
ment. Some studies (e. , Noterdaeme , Amorosa , Mildenberger , Sitter , & Minow
2001; Ozonoff & Jensen , 1999) have considered how various executive components
are related to different developmental disorders (e. , autism vs. attention-
deficit/hyperactivity disorder), while others have investigated how the cognitive
processes encompassed by the executive dysfunction hypothesis relate to autism. For
example , Russell and his colleagues have shown that children with autism struggle
with (1) disengaging attention from a focal object (Hughes & Russell , 1993), (2) atten-
tional set- shifting and planning (Hughes , Russell , & Robbins , 1994), (3) choosing
between equally weighted alternatives (Russell , Jarrold , & Henry, 1996), (4) encod-
'Unlike Ozonoff , Pennington , and Rogers (1991) or Pennington et al. (1997), we do not consider exec-
utive dysfunction to be the primary cause of autism. We do , however , believe executive dysfunction to be
an important aspect of autism , functioning as a contributing factor but not as a causal agent. Thus , like Liss
et al. (2001), we maintain that " although impaired executive dysfunction is a commonly associated feature
of autism... (it) is unlikely to cause (autism)" (p. 261).
, p.
037- 056 10/16/03 5 : 09 Page 40
177235-
ing rules in a verbal form (Russell Jarrold , & Hood , 1999), and (5) following arbitrary
procedures (Biro & Russell , 2001).
Problem solving inevitably requires an intentional focus on some parts of a prob-
lem display, some aspects of an associated mental representation , or some portion of
a knowledge base. Identifying the details relevant to a problem entails excluding
irrelevant ones that may appear to be connected but are actually misleading.
Attention to these irrelevant cues must be inhibited. As Chao and Knight (1997
67) point out: " One of the main functions of the attentional system is to enable cog-
nitive processing to focus on stimulus attributes and ignore or inhibit irrelevant
aspects. " The ability to control attention and inhibition in this matter develops sig-
nificantly in the first five years (Diamond , 2002; Diamond & Taylor , 1996; Gerstad
Hong, & Diamond , 1994) and has been credited for a variety of important changes
in problem solving (Dagenbach & Carr , 1994; Dempster , 1992; Harnishfeger &
Bjorklund , 1993).
The dimensional change card sort task , developed by Zelazo and his colleagues
(e. , Zelazo & Frye , 1997) and administered to children with autism (Zelazo et al.
2002) provides a means of examining the development of these processes. Children
are required to sort a set of cards by one dimension and then to re- sort the same cards
by a competing dimension. Thus , the feature that is attended to in the first stage must
be ignored in the second. This is difficult for typically- developing children and they
are not successful until they are about 5-years old. The error is to continue to sort the
cards according to the pre- switch rule.
We explored the problem in more detail by creating a computerized version of
the dimensional change card sort task consisting of four conditions , called (1) the
colour game , (2) the colour- shape game , (3) the colour- object game , and (4) the func-
tion- location game (Bialystok & Martin , in press). The intention was to vary the atten-
tion and inhibition demands of each condition to isolate their role in solving the task.
All four conditions required the cards to be reclassified in the post- switch phase , but
all four presented different representational (and , therefore , inhibition) problems.
One of the conditions , the colour game, was based on a single perceptual feature; two
conditions , colour- shape and colour- object, were based on two perceptual features
and the fourth condition , function- location , was based on a conceptual feature of the
depicted objects. The task difficulty is in re- assigning the stimuli to a new perceptual
(or conceptual) category when there is interference from the previously relevant cat-
egory that needs to be ignored (Bialystok & Martin , in press).
The task involves two kinds of inhibition , and the results point out the difference
between them. The first response inhibition is the suppression of a practised response
or association between a stimulus and response that needs to be overridden. In a study
by Jacques , Zelazo , Kirkham , and Semcesen (1999), children observed as a puppet
performed the classification and the child simply decided if the puppet sorted cor-
rectly or not. This procedure eliminated the need for children to carry out the motor
response and therefore eliminated the need to inhibit this response in the post- switch
phase. Nonetheless , performance in this modified task was no better than that on the
standard version and they concluded that inhibition was not a factor in solving the
037- 056 10/16/03 5 : 09 Page 41
177235-
problem. Similarly, Perner (2001) demonstrated that a control condition that requires
reversing the destination of the card following the switch , even in the absence of con-
flicting information from the targets , does not improve children s performance. This
is the kind of inhibition required in the colour game.
The second conceptual inhibition, is the suppression of attention to a salient but
competing mental representation. This is the inhibition central to the other three
conditions- attention has to be shifted from one stimulus feature (e. , colour) to
another (e. , shape), even though the initial feature (colour) remains present and
salient. The function- location condition is more difficult than the other two because
the classifications are not based directly on perceptual features-children need to
identify the objects and then classify them to one of the two categories (Bialystok &
Martin , in press). At the same time , the classification itself is difficult, and children
make many errors in both the pre- switch and post- switch phases. This condition
therefore , requires higher levels of representation ability than the other three to iden-
tify, classify, and then re- classify the stimuli.
The interpretation that conceptual inhibition is responsible for problem difficulty
is supported by research that has assisted children s focus on the relevant dimension
in the post- switch phase. This can be achieved by removing one of the stimulus fea-
tures during the post- switch phase (Bialystok & Martin , in press; Zelazo & Frye , 1997;
Zelazo & Jacques , 1996), questioning children about the feature to which they are
attending (Towse , Redbond , Houston- Price , & Cook , 2000), or requiring children to
label the relevant dimension on each sorting trial (Kirkham , Cruess , & Diamond , in
press). These methods all significantly improve performance and point to the role
conceptual inhibition as the primary difficulty in the task.
In sum , the dimensional change card sort task includes at least three kinds of pro-
cessing demands: response inhibition , conceptual inhibition , and representational
complexity. The potential to isolate these three processes makes the task a promising
test of the kind of executive process that might be impaired in children with autism.
If children with autism have a specific impairment in the ability to inhibit attention
to prepotent responses (response inhibition) or to competing mental representations
(conceptual inhibition), then they wil find the task difficult. Comparing their per-
formance to the different conditions of this problem and to the performance of chil-
dren who are comparable in verbal mental age and chronological age wil help to
identify one of the core cognitive deficits in autism. The hypothesis was that children
with autism would be delayed in their ability to solve the problems requiring con-
ceptual inhibition and that increasing demands on representational complexity
would disproportionately disrupt their performance relative to the typically- devel-
oping children. In contrast , we expected no relative deficit in their performance on
the colour game that depends primarily on response inhibition.
037- 056 10/16/03 5 : 09 Page 42
177235-
Method
Participants
This study compared a group of children with autism and two groups of typically-
developing children on their ability to solve the four conditions of the dimension
change card sort task. The first group included 15 children (14 boys and 1 girl; mean
age = 8;4 years) who had received a diagnosis of Autistic Disorder (American
Psychiatric Association , 1994). The cognitively- matched control group included 184-
year- olds (9 boys and 9 girls; mean age = 4;7 years) and the age- matched control
group included 11 8-year- olds (5 boys and 6 girls; mean age = 8;0 years).
Parents , educators , and clinicians verified diagnoses for children with Autistic
Disorder and relevant records were consulted whenever possible. Participants were
recruited through announcements to service organizations , day care centers , public
schools , and special programs in major urban centres. Participants ' socio- economic
status ranged from middle class to upper-middle class. All participants had received
their diagnosis of Autistic Disorder from qualified medical experts according to
DSM- IV (American Psychiatric Association , 1994) diagnostic criteria. Participants
were free from psychoactive medication and no participants were diagnosed with co-
morbid mental disorders.
Instruments
Forward Digit Span. The forward digit span was used to assess verbal short- term
memory capacity (WISC- R; Weschler , 1974). A sequence ofrandom digits was read
twice , and the child was asked to repeat this sequence to the experimenter. The test
began with a sequence of three digits , and the number of digits increased by one after
each correct trial. If an error occurred, a second sequence was given at the same
length. Testing ended when two consecutive errors were made at the same length.
The number of digits in the last correct sequence is the forward digit span score. One
child with autism was unable to perform the forward digit span.
Ravens Coloured Progressive Matrices (sets A , AB , B). The Coloured Progressive
Matrices (Raven , Court , & Raven , 1986) was employed to assess visual- spatial abili-
ties (i. , abilities to arrange spatial perceptions into semantically related wholes) and
to assess comparison- making skils. The task is comprised of 36 pictures (organized
as 3 sets of 12 items), each of which has a portion missing. Children must select which
of six possible segments correctly completes the picture. The score is the total num-
ber of items out of 36 that the child completes correctly. One child with autism was
unable to complete the task.
Peabody Picture Vocabulary- Revised (PPVT- R). This task was used to assess
receptive vocabulary and verbal mental age (Dunn & Dunn , 1981). The child is asked
to select one picture from a set of four that ilustrates the word spoken by the exper-
037- 056 10/16/03 5 : 09 Page 43
177235-
imenter. Items become more difficult as testing progresses. Testing continues until the
participant makes six mistakes in eight consecutive responses. Raw scores are con-
verted to standard scores by means of normalized tables.
Computerized Dimensional Change Card Sort Task. This task included four con-
ditions (games) that were presented to the children on an IBM Thinkpad. A black
cover was placed over the keyboard exposing only three keys: the " , the " , and
the spacebar. The " W" and " P" keys were selected for their symmetry to the screen
and were covered with keycaps clearly marked " X" and " , respectively. The
spacebar was covered with a blank white keycap. An animated character appeared
on the screen before each new game and each switch phase in order to explain the
rules of each game to the child. For each game , the stimulus to be classified appeared
in the centre of the screen and children were given instructions about which key to
press. For three of the games , small target stimuli also appeared near the bottom of
the screen directly aligned with one of the response keys. This configuration is ilus-
trated in Figure 1 for each of the four conditions. The child was allowed three prac-
tice trials with feedback. If the child was incorrect on one of the practice trials , the
rules of the game repeated automatically. This continued until the child was able to
perform the three practice trials accurately. The experimenter remained with the
child and repeated instructions if the child did not fully understand the task.
OR8
a. Colour Game
The children were told to watch for either a blue square or a red square. In the
first set of rules , they were told to press the button with the X on it when the red
square appeared , and to press the button with the 0 on it when the blue square
appeared. There were 10 items in this set , including equal numbers of red and blue
squares presented in a random order. Following this , they were told that the rules had
now switched and they were to press the X if the blue square appeared , and the 0 if
the red square appeared. Again , there were 10 trials , and the squares were presented
in a random order. The order of these rules was counterbalanced across all children.
This condition , consisting of red and blue circles and squares , was presented as
the colour game or the shape game in counterbalanced order. If the colour game was
first , the child was told to put all the blue pictures in the box with the blue picture on
, and all the red pictures in the box with the red picture on it. Two boxes appeared
at the bottom of the screen with the appropriate picture on the box, and the image
to be sorted appeared in the middle of the screen. If the boxes had a red square and
a blue circle on them , then the pictures to be sorted were red circles and blue squares.
The child pressed the button closest to the box that the picture was to be placed in.
The picture visually dropped into the appropriate box. When the first part of the
game was completed , the animated figure appeared again and explained the post-
switch rules. In this example , the new game was the shape game, and children were
told to place the squares in the box with the square on it , and the circles into the box
with the circle on it. Again , there were 10 trials in each of the two games , consisting
of five circles and five squares presented in a random order. The programme ran-
domized whether the sorting stimuli were blue circles and red squares or red circles
and blue squares.
c. Colour-Picture Game
This task was identical to the previous condition except that the stimuli were red
or blue rabbits or flowers.
d. Function-Location Game
Unlike the previous three conditions , the sorting dimensions in this game were
not perceptual features but functional properties of the stimuli. In the first part of the
"We did not expect any difference between these two conditions but wanted to vary the kind of stim-
uli that was used.
p;: .
037- 056 10/16/03 5 : 09 Page 45
177235-
game , the children had to sort the pictures based on whether they would play with
the item or wear the item. All 10 pictures clearly belonged to one category or the
other (these were: a bicycle , a skateboard , a pail and shovel , a skipping rope , a kite
a pair of slippers , a nightgown , a bib , ballet slippers , and baby pyjamas). The post-
switch rules required the child to sort the pictures based on whether these items
belonged inside or outside the house. The target stimuli on the sorting compartments
were a winter jacket (wear , outside) and a teddy bear (play, inside). There were
items in each of the two games and the order (function and location) was counter-
balnaced across subjects.
The task demands change across these conditions. The colour game requires
response inhibition to resist placing the card in the previously correct manner but no
conceptual inhibition because there is only one possible dimension to which to
attend , namely colour. Representing this rule is also straightforward. The colour-
shape and colour object games require response inhibition because , like the colour
game , each card must be placed in the box opposite to that used in the pre- switch
phase. Additionally, however , these conditions require conceptual inhibition because
the dimension that previously defined the correct placement is stil present but is now
misleading. The representation demands require the ability to represent two percep-
tual dimensions. The function- location game has the same demands for response
inhibition and conceptual inhibition as do the two previous games , but the repre-
sentational demands are more difficult because the classification needed is based on
functional properties that must be conceptualized from the pictures rather simple
perceptual attributes.
The conditions were presented in the fixed order described above but items were
randomized within the games. The order of games (e. , colour versus shape) within
conditions was counterbalanced. The scores were the number of correct classifica-
tions out of 10 in each of the pre- switch and post- switch phases of each game.
Procedures
The children were tested individually during two separate sessions that lasted
roughly 25 to 35 minutes. Testing sessions were self- paced; breaks were permitted to
ensure attention and co- operation. During Session 1 , participants received the for-
ward digit span , the Ravens Matrices and the PPVT- R. During Session 2 , participants
completed the computerized dimensional change card sort task.
Results
Children s scores on the background cognitive measures are reported in Table
A one-way ANOVA for digit span scores indicated that the difference between the
groups was not significant F(2 42) = 2.
~. ~. ~.
037- 056 10/16/03 5 : 09 Page 46
177235-
PPVT raw scores and standard scores are presented in Table 1. PPVT raw scores
reflect absolute performance levels. Although PPVT standard scores (i. , the nor-
malized calculation of PPVT raw scores in terms of the child' s age) are generally
more meaningful because they allow children to be compared across ages , they are
not utilized in this study (but are reported in the table) because a standardized con-
version according to chronological age is inappropriate for children with autism.
Predictably, a one- way ANOVA for group on these standard scores indicated a sig-
nificant difference 42) = 73. 0001 , and Duncan contrasts revealed that
the 4- year- olds and 8 year olds levels were not significantly different from each other
but the children with autism scored significantly lower than the other two groups.
This is the expected finding, because the purpose of the standard scores is to adjust
the raw scores according to age of typically- developing children , making them all
equivalent irrespective of age.
A one- way ANOVA on PPVT raw scores was used to determine the relative pro-
ficiency of the autistic and typically- developing groups without considering age dif-
ferences. This analysis showed an effect of group, 43) = 35. 0001 , and
Duncan contrasts revealed three distinct levels. Therefore , the receptive vocabulary
of the children with autism is more impoverished than that of typically developing 4-
year- olds in absolute (raw) terms.
A one-way ANOVA for the total score on the Ravens Matrices revealed an effect
of group, , 42) = 9.45 0004 , caused by the difference between the 8-year-
olds who scored the highest and the 4- year- olds who scored the lowest. These scores
are reported in Table 1. The children with autism performed the same as typically-
developing children on the Ravens Progessive Matices , a test of nonverbal intelli-
gence , but were significantly lower than the two control groups in a test of receptive
vocabulary knowledge (PPVT).
Results for the card sort task are analysed in two ways. The traditional procedure
is to determine a criterion level of performance and classify children as either pass-
ing or failing the post- switch phase (Zelazo & Frye , 1997). The present analyses use
both the conventional pass- fail categories and a more detailed three- way classifica-
tion that includes a category for guessing. Second , to obtain a more graded under-
standing of the results , analysis of variance and covariance was used to determine the
effect of group and task differences on performance.
Children were classified into one of three categories based on their score in the
post- switch phase of each condition. Children who correctly sorted between 7 and 10
4Another control group consisting of 5-year- olds , not reported in this paper , scored exactly the same
as the children with autism on the Ravens Matrices but was even higher than the 4-year- olds included here
on the PPVT. Therefore , the 4- year- olds were considered to be a more appropriate control group.
~. =. -( .
items in the post- switch phase of a particular game were classified as " Correct"
responders for that game; children who sorted between 4 and 6 items correctly were
called " Chance " responders; and those who correctly classified only 0 to 3 items in
the post- switchwere designated as " Perseverators " for that game. This adaptation of
the usual binary classification is more precise because it enables a distinction to be
made between children who fail because they perseverate and those who fail because
they respond at chance. The combination of these two non- correct categories yields
the usual pass- fail distinction. The number of children in each category for each of
the games is reported in Table 2.
The chi- square analyses , conducted for both the two- way and three-way classifi-
cation of responders , indicate the extent to which the distribution across the cate-
gories is the same for the three groups. The results are indicated in the table beside
the distribution for each condition. Chance responding was rare , and the consistency
of the analysis for both the two- and three- category classification suggests that guess-
ing was not a problem , or at least , not a problem specific to only one of the groups.
The only distribution to differ from chance is the colour condition in which the chil-
dren with autism were disproportionately weak. For the other three conditions , the
proportion of children passing and failing was the same. The result does not change
if the chi- square analysis compares only the children with autism and the 4-year- olds
for their likelihood of passing and failing in each condition. Again , the distribution is
different only for the colour game , X' (1)= 6. 01. Therefore , the responses of
the children with autism are qualitatively different from those in the other two groups
for the colour condition but the same as the other children for the other three.
Although children passed the other three games in equivalent proportions , their
relative success on these games may stil be different. Therefore , analysis of variance
was used to provide more detail about children s performance. The mean scores out
of 10 for the pre- and post- switch phases of each condition are displayed in Figures
2 and 3 respectively. A 3- way ANOVA for group (3), condition (4), and phase (2)
Table 2: Distrbution
of responses by group on the
computeried dimensional change card sort tak
Game Group Correct Fail Chance Perseverator Pass- Fail 2 3- Category
(Pass) df= 2 df= 4
indicated main effects of group, F(2 41) = 18.10 0001 , MSE = 11.39 , condition
F(3 123) 0003 , MSE = 5. , and phase F(1 41) = 78.12 0001 , MSE=
12. 27. The group effect was that all three groups were significantly different from each
other , with the highest scores obtained by the 8- year- olds and the lowest by the chil-
dren with autism. The condition effect indicated that the colour condition was easier
than the other three , which did not differ from each other. The phase effect reflected
the higher performance on the pre- switch problems.
The children with autism obtained lower scores than the other children in all four
conditions. However , the children with autism also had lower receptive vocabulary
scores than the other children , and this disadvantage in verbal skil may have been
partly responsible for their performance. Therefore , scores were re- examined using
EJ 8 years
4 years
Autism
8 years
4 years
Au ti sm
ANCOVA analysis with PPVT Stanine scores as the covariate to establish perform-
ance levels that might be obtained assuming relative equivalence of vocabulary lev-
els. The least squares means generated from this analysis for each group and
condition are presented in Table 3. Note that these means do not correspond to those
plotted in Figures 2 and 3 because they have been adjusted for vocabulary ability.
Because the executive processing demands of each condition were different , sep-
arate analyses were conducted for each to focus on group differences in performance
of these specific problems when vocabulary knowledge has been controlled.
Consequently, there were four two-way ANCOVA (PPVT covariate) analyses for
phase (pre- and post- switch) and group. For the colour game , there was an effect of
group, , 39) = 6. 003 , MSE = 4. , in which all three groups were differ-
ent from each other , the 8- year- olds achieving the highest and the children with
autism receiving the lowest scores. The colour- shape game also revealed an effect of
group, 39) = 3. , MSE = 7.88 , this time indicating that both the 8- year-
olds and 4- year- olds (not differing from each other) performed better than the chil-
dren with autism. The colour- object game produced only a main effect of phase
39) = 13.14 0008 , MSE = 10.17 , in which pre- switch scores were consistently
higher than post- switch for all the children. Finally, the function- location condition
revealed an effect of group, F(2 39) = 4. , MSE = 7.96 , and phase F(1 39)
= 7.27 01. The 4-year- olds obtained the lowest scores with no difference between
the 8- year- olds and children with autism. Pre- switch trials were reliably better than
post- switch trials for all the children.
Discussion
A group of 8- year- old children with autism was compared to a group of typically-
developing 4-year- olds who were expected to match for mental age and to a group
of typically- developing 8- year- olds who were matched for chronological age on a
computerized version of the dimensional change card sort task. All the children per-
formed similarly on a digit span test, a rough guide to verbal short- term memory
capacity. The children with autism were not significantly different from either the 4-
year- olds or the 8- year- olds on the Ravens Progressive Matrices , although these two
groups were different from each other. Scores for receptive vocabulary, however
indicated an important disparity between the groups. This is an unfortunate outcome
but it may be unavoidable. The standardized receptive vocabulary scores for the
Table 3: Least squares means (PPVT Covarate) and stadard errors on the
computeried dimensional change card sort tak by group
Group Colour Colour- Colour- ect F unction- Location
Pre Post Pre Post Pre Post Pre Post
Autism 9 (. 71) 5 (1.2) 2 (. 87) 2 (1.7) 6 (.63) 2 (1.7) 8.4 (. 75) 5 (1. 7)
years 8.4 (. 52) 7.1 (.85) 8 (. 63) 5.4 (1.2) 9.4 (.46) 7 (1.3) 6 (. 55) 3 (1.2)
years 6 (.49) 5 (. 82) 9.1 (. 60) 7.4 (1.2) 8 (.44) 5.4 (1.2) 7 (.52) 2 (1.2)
037- 056 10/16/03 5 : 09 Page 50
177235-
year- old children with autism was lower than that for typically- developing 4- year-
olds , but matching to a control group younger than 4-years old would not have been
reasonable for the other tasks. A group of 3- year- olds , for example , may have pro-
vided a better match in receptive vocabulary for the children with autism ,
but they
would not have provided a match on the Ravens Progressive Matrices and their per-
formance on the experimental sorting task would likely have been unreliable , if they
could do it at all. Our solution , therefore , was to use analysis of covariance to statis-
tically control for initial differences in levels of vocabulary knowledge.
The data were examined by means of chi- square analysis , ANOVA , and
ANCOVA to investigate different aspects of the results. First, the chi square analyses
compared the proportion of children who passed or failed each of the conditions
across groups. This proportion was only different for the colour game , where a sig-
nificantly larger group of children with autism failed the post- switch phase. This pat-
tern indicates a qualitative difference in the way in which children in these three
groups solve this problem. Importantly, the chi- square analysis of the three-way clas-
sification indicated that very few of the children were chance responders , although
the majority of guessers were children with autism.
Second , the ANOVA analysis showed that the scores of the children with autism
were lower than those of all the other children across all the conditions. The main
effect of group also indicated the superiority of the 8- year- olds for all the conditions.
This was a difficult task , especially for the children with autism.
Finally, the ANCOVA provides more detailed understanding by comparing the
performance of each group on each condition controlling for initial differences in
vocabulary knowledge. These results showed that the children with autism per-
formed lower than the other children in both the colour game and the colour- shape
game , but were not lower on the other two games. In fact , these children were better
than the 4- year- olds in the function- location condition. Once corrections for vocab-
ulary knowledge have been made , the children with autism were only slightly disad-
vantaged on these semantically complex conditions based on more sophisticated
representations. Therefore, initial deficits in language proficiency account for some
but not all of the diffculty that children with autism experience on this task.
Evidence that children with autism score poorly relative to typically- developing
children on a cognitive task is not surprising; it is more important to identify the spe-
cific deficit responsible for that poor performance. Consider the three processes
involved in solving this task: representation , response inhibition , and conceptual
inhibition. The 4-year- olds experienced difficulties with representation and were par-
ticularly challenged by the function- location game; the children with autism experi-
enced difficulties with response inhibition even though the colour game was easy for
the other children; and all the children , including the 8-year- olds , found the concep-
tual inhibition demands to be challenging, as even this oldest group did not achieve
ceiling performance on the most difficult condition.
Three aspects of this pattern are striking. First , children with autism are distin-
guished by their difficulty in inhibiting a familiar or learned response association to
a single perceptual cue as in the colour game. This finding is congruent with a grow-
037- 056 10/16/03 5 : 09 Page 51
177235-
ing body of research (e. , Bryson , Landry, & Wainwright , 1997) implicating execu-
tive function deficits , especially inhibitory control (e. , Hughes & Russell , 1993), in
the pathogenesis of autism. Second , when initial vocabulary levels were controlled
children with autism performed as well as typically- developing 8-year- olds and sig-
nificantly better than typically- developing 4- year- olds on a problem that was based
on a complex conceptual classification. Hence , the ability of children with autism to
represent conceptual information , once vocabulary had been adjusted , was not
impaired. This optimistic interpretation is confirmed by their performance on the
Raven s Matrices. Together , these results locate the source of difficulty for the chil-
dren with autism in the need for inhibition , particularly for response inhibition.
Finally, the surprising finding that typically- developing 8- year- olds continue to per-
severate on the dimensional change card sort task , especially in its more complex
representational forms , suggests that the development of inhibitory executive
processes is more protracted than previously believed (e. , Zelazo , Frye , & Rapus
1996).
In the executive dysfunction perspective on autism , inhibitory control is a pri-
mary source of cognitive impairment (e. , Hughes & Russell , 1993). Our results cor-
roborate this view but point to a particular deficit in response inhibition that is not
evident in the other typically- developing children. Children with autism also have
less control over conceptual inhibition than do typically- developing children , but
only when the conceptual basis for the classification is easy (colour- shape game). As
that conceptual basis becomes slightly richer , requiring more detailed mental repre-
sentations , the children with autism do just as well as the other children if language
proficiency is taken into account (colour- object game) and better than 4-year- olds if
the classification is based on functional categories instead of perceptual ones (func-
tion- location game). Future efforts aimed at specifying the exact inhibitory executive
processes impaired in children with autism may want to distinguish between
response inhibition and conceptual inhibition so as to enable precise conclusions.
The results point to an executive deficit in children with autism in which they
have difficulty in disengaging from a simple ongoing cognitive routine and making
an adjustment in that routine. Children with autism are compromised both in the
extent to which inhibition processes are functional and in the nature of inhibition
processes that are impaired. These children not only experienced deficits in per-
formance on all tasks requiring conceptual inhibition , but also experienced deficits
in a condition requiring response inhibition. These inhibition processes are central to
the executive system that is responsible for planful and controlled problem solving.
Such executive function deficits may contribute to social communication impair-
ments in children with autism. Unable to flexibly and efficiently disengage and coor-
dinate attention , the rapid shifts and constant changes that typify social
communication confuse children with autism (e. , Courchesne et al. , 1999; McEvoy
et al. , 1993).
(i. , cognitive deficits) of the symptomatology of children with autism , they do not
explain the potentially simultaneous involvement of other impairments in different
domains. It may be that cognitive deficits (such as inhibition of attention) underlie
social difficulties (such as mindblindness) or that social difficulties encourage cogni-
tive deficits or that cognitive deficits and social difficulties emerge in a complex
dialectical transaction (Fine , Lumsden , & Blair , 2001). Clearly, no single factor or uni-
tary theory is likely to adequately explain the severity and complexity of autism.
References
Adrien j. L. , Martineau j., Barthelemy, C. , Bruneau , N. , Garreau , B. , & Sauvage , D. (1995).
Disorders of regulation of cognitive activity in autistic children. Journal of Autism and
Developmental Disorders , 294-296.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders
(4th ed. ). Washington , DC: American Psychiatric Association.
Baron- Cohen , S. (1989). The autistic child' s theory of mind: A case of specific developmental
delay. Journal of Child Psychology and Psychiatry and Allied Disciplines , 285-297.
Baron- Cohen , S. (1994). From attention- goal psychology to belief- desire psychology: The
development of a theory of mind , and its dysfunction. In S. Baron- Cohen , H. Tager-
Flusberg, & D. j. Cohen (Eds. Understanding other minds: Perspectives from autism
(pp. 59- 82). New York: Oxford University Press.
Baron- Cohen , S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge , MA:
The MIT Press.
Baron- Cohen , S., Leslie , AM. , & Frith , U. (1985). Does the autistic child have a " theory of
mind" Cognition , 37-46.
Bennetto , L. , Pennington , B. F. , & Rogers , S. j. (1996). Intact and impaired memory functions
in autism. Child Development , 1816- 1835.
Bialystok , E. , & Martin , M. M. (in press). Attention and inhibition in bilingual children:
Evidence from the dimensional change card sort task. Developmental Science.
Biro , S. , & Russell , j. (2001). The execution of arbitrary procedures by children with autism.
Development and Psychopathology, , 97- 110.
Boucher , j. (1981). Memory for recent events in autistic children. Journal of Autism and
Developmental Disorders , 293- 301.
Boucher , j., & Lewis V. (1989). Memory impairments and communication in relatively able
autistic children. Journal of Child Psychology and Psychiatry and Allied Disciplines, 30
99- 122.
Bryson , S. E. , Landry, R , & Wainwright j. A. (1997). A componential view of executive dys-
function in autism: Review of recent evidence. In Burack , j.A. , & Enns , j.T (Eds.
Attention, development, (pp. 232- 259). New York: The Guilford Press.
and psychopathology,
Chao , L.I. , & Knight, RT (1997). Prefrontal deficits in attention and inhibitory control with
aging. Cerebral Cortex , 63- 69.
Courchesne , E. , Townsend j., Akshoomoff, N. A , Saitoh , 0. , Yeung- Courchesne , R , Lincoln
j., et al. (1999). Impairment in shifting attention in autistic and cerebellar patients.
Behavioral Neuroscience 108 , 848- 865.
Dagenbach , D. , & Carr , T H. (1994). Inhibitory processes in perceptual recognition: Evidence
for a center- surround attentional mechanism. In Dagenbach , D. , & Carr , TH. (Eds.
Inhibitory processes in attention, memory, and language (pp. 327-357). San Diego , CA:
Academic Press.
Dahlgren , S. 0. , & Trilingsgaard , A (1996). Theory of mind in non- retarded children with
autism and Asperger s syndrome: A research note. Journal of Child Psychology and
Psychiatry and Allied Disciplines , 759- 763.
, j. ), ),
Dempster , F. N. (1992). The rise and fall of the inhibitory mechanism: Toward a unified the-
ory of cognitive development and aging. Developmental Review , 45- 75.
De Viliers (1999). Language and theory of mind: What are the developmental relation-
ships? In S. Baron- Cohen , H. Tager- Flusberg, & DJ. Cohen (Eds. Understanding other
minds: Perspectives from autism and cognitive neuroscience (pp. 83- 123). Oxford: Oxford , UK:
University Press.
Diamond , A. (2002). Normal development of prefrontal cortex from birth to young adulthood:
Cognitive functions , anatomy, and biochemistry. In D. T. Stuss & RT. Knight (Eds.
Principles of frontal lobe function (pp. 466- 503). London: Oxford University Press.
Kirkham , N. , Cruess , L , & Diamond , A. (in press). Helping children apply their knowledge to
their behavior on a dimension- switching task. Cognition.
Leekam , S. R , & Pemer j. (1991). Does the autistic child have a metarepresentational deficit?
Cognition , 203-218.
Liss , M. , Fein , D. , Allen , D. , Dunn , M. , Feinstein , C. , Morris , R , et al. (2001). Executive func-
tioning in high- functioning children with autism. Journal of Child Psychology and Psychiatry
and Allied Disciplines , 261-270.
McEvoy, R E. , Rogers , S. j., & Pennington , B. F. (1993). Executive function and social com-
munication deficits in young autistic children. Journal of Child Psychology and Psychiatry and
Allied Disciplines , 563-578.
Noterdaeme , M. , Amorosa , H. , Mildenberger , K , Sitter , S. , & Minow , F. (2001). Evaluation of
attention problems in children with autism and children with a specific language disor-
der. European Child Adolescent Psychiatry, , 58- 66.
, M. , & Yirmiya , N. (1993). Autism and theory of mind. Psychologia: Israel Journal
Psychology, , 126- 139.
Ozonoff, S. (1995). Reliability and validity of the Wisconsin Card Sorting Test in studies of
autism. Neuropsychology, , 491-500.
Ozonoff, S. , &Jensen j. (1999). Brief report: Specific executive function profies in three neu-
rodevelopmental disorders. Journal of Autism and Developmental Disorders , 171- 177
Ozonoff, S. , & McEvoy, R E. (1994). A longitudinal study of executive function and theory of
mind development in autism. Development and Psychopathology, , 415-431.
Ozonoff, S. , Pennington , B. F. , & Rogers , S. j. (1991). Executive function deficits in high- func-
tioning autistic individuals: Relationship to theory of mind. Journal of Child Psychology and
Psychiatry and Allied Disciplines , 1081- 1105.
Ozonoff, S. , & Strayer , D. L. (1997). Inhibitory function in nonretarded children with autism.
Journal of Autism and Developmental Disorders , 59-
Pelios , L. v. , & Lund , S. K (2001). A selective overview of issues on classification , causation
and early intensive behavioral intervention for autism. Behavior Modifcation , 678- 697.
Pennington , B. F. , & Ozonoff, S. (1996). Executive functions and developmental psy-
chopathology. , 51- 87.
Journal of Child Psychology and Psychiatry and Allied Disciplines
Pennington , B. F. j., Bennetto , L. , Griffith , E. M. , Reed , D. T. , & Shyu, V. (1997).
, Rogers , S.
Validity tests of the executive dysfunction hypothesis of autism. In j. Russell (Ed.
Autism as an executive disorder (pp. 143- 173). New York: Oxford University Press.
Pemer (2001). Paper presented in symposium Rule use through the lens of the dimensional
change card sort: What develops?" at the biennial meeting of the Cognitive Development
Society, Virginia Beach , VA , October 26- 27.
Pemer , j., Frith , U. , Leslie , A. M. , & Leekam , S. R (1989). Exploration of the autistic child'
theory of mind: Knowledge , belief, and communication. Child Development , 686- 700.
Peterson , C. C. & Siegal , M. (1995). Deafness , conversation and theory of mind. Journal of Child
Psychology and Psychiatry and Allied Disciplines , 459-474.
Prior , M. , Dahlstrom , B. , & Squires , T.L. (1990). Autistic children s knowledge of thinking and
feeling states in other people. Journal of Child Psychology and Psychiatry and Allied
Disciplines , 587- 601.
Prior , M. , & Hoffman , W. (1990). Brief report: Neuropsychological testing of autistic children
through an exploration with frontal lobe tests. Journal of Autism and Developmental
Disorders , 581-589.
Raven j.C. , Court j.H. , & Raven j. (1986). Coloured Progressive Matrices. London: H. K Lewis.
Reed , T. & Peterson , C. C. (1990). A comparative study of autistic subjects ' performance at two
levels of visual and cognitive perspective taking. Journal of Autism and Developmental
Disorders , 555-567.
Rumsey, j. M. (1985). Conceptual problem- solving in highly verbal , nonretarded autistic men.
Journal of Autism and Developmental Disorders , 23-36.
Rumsey, j. M. , & Hamburger , S. D. (1988). Neuropsychological findings in high- functioning
men with infantile autism , residual state. Journal of Clinical and Experimental
Neuropsychology, 201- 221.
j., ), ),
Russell Jarrold , c. , & Henry, L. (1996). Working memory in children with autism and mod-
erate learning difficulties. Journal of Child Psychology and Psychiatry and Allied Disciplines
, 673- 686.
Russell , j., Jarrold , C. , & Hood , B. (1999). Two intact executive capacities in children with
autism: Implications for the core executive dysfunctions in the disorder. Journal of Autism
and Developmental Disorders , 103- 112.
Shu , B. , Lung, F. W , Tien , A. Y , & Chen , B. C. (2001). Executive function deficits in non-
retarded autistic children. Autism , 165- 174.
Szatmari , P. , Tuff, L. , Finlayson , A. j., & Bartolucci , G. (1990). Asperger s syndrome and
autism: Neurocognitive aspects. Journal of the American Academy of Child and Adolescent
Psychiatry, , 130- 136.
Towse , j.N. , Redbond , j., Houston- Price , C. , & Cook , S. (2000). Understanding the
dimensional change card sort: Perspectives from task success and failure. Cognitive
Development , 325- 343.
We schier , D. A. (1974). Intelligence Scale for Children-Revised.
Wechsler San Antonio , TX:
Psychological Corporation.
Wimmer , H. , & Perner j. (1983). Beliefs about beliefs: Representation and constraining func-
tion of wrong beliefs in young children s understanding of deception. Cognition, 13
103- 128.
Yirmiya , N. , Erel , 0. , Shaked , M. , & Solomonica- Levi , D. (1998). Meta- analyses comparing
theory of mind abilities of individuals with autism , individuals with mental retardation
and normally developing individuals. Psychological Bulletin 124 , 283- 30Z
Zelazo , P. D. , & Frye , D. (1997). Cognitive complexity and control: A theory of the develop-
ment of deliberate reasoning and intentional action. In M. Stamenov (Ed. Language
structure, discourse, and the access to consciousness (pp. 113- 153). Amsterdam: John
Benjamins.
Zelazo , P. D. , Frye , D. , & Rapus , T. (1996). An age- related dissociation between knowing rules
and using them. Cognitive Development , 37- 63.
Zelazo , P. , &Jacques , S. (1996). Children s rule use: Representation , reflection and cogni-
tive control. In R. Vasta (Ed. Annals of child development (Vol. 12 , pp. 119- 176). London:
Jessica Kingsley.
Zelazo , P. Jacques , S. , Burack j.A. , & Frye , D. (2002). The relation between theory of mind
and rule use: Evidence from persons with autism- spectrum disorders. Infant and Child
Development , 171- 195.
Ellen Bialystok
Department of Psychology
York University
4700 Keele Street
Toronto, Ontario, M3J
Canada
Email: ellenbr:yorku.
Abstract. The neuropathology of autism primarily involves limbic system and cerebellar
structures. The neurons of the dentate gyrus and stratum pyramidale in the hippocampal for-
mation are smaller in size and are densely packed and some neurons in the CA sub fields, have
reduced dendritic arbors. Research in our laboratory has ftcused on the GABAergic system in
autism due to initial findings of a statistically signifcant reduction in the density of tHJ-
flunitrazepam labeled benzodiazePine binding sites and of tHJ- muscimol labeled GABA
receptors in high binding regions in the hippocampus. The ftrmer study was ftllowed up with
a multiple concentration saturation binding study demonstrating a trend for a reduction in the
number of benzodiazePine binding sites (change in the Bmax) but normal binding affnity (no
change in the Kd). These results suggest possible changes in the modulation ofGABA receptors
in the hiPpocampus and, that there may be more widespread effcts of altered GABAergic func-
tion in the autistic brain.
Introduction
Consistent neuropathological findings in autism directly implicate limbic struc-
tures including the hippocampal formation , amygdala and basal forebrain and
cerebellar structures including the cerebellar cortex , cerebellar nuclei and the infe-
rior olivary complex (Bauman and Kemper , 1985 , 1994). Smaller cell size and
increased cell packing density is found in the hippocampus and in the medial part of
the amygdala whereas age- related changes (larger than normal size neurons in chil-
dren; smaller than normal size neurons in adult) are found in the nucleus of the
Diagonal band of Broca in the basal forebrain , cerebellar nuclei and the inferior oli-
vary complex (Bauman and Kemper , 1985). Using the Golgi method , pyramidal cells
from the CAI and CA4 hippocampal fields showed reduced complexity and extent
of the dendritic arbors (Raymond et al. , 1996).
Evidence that the GABAergic system is affected in autism emerged with the find-
ing of decreased numbers of Purkinje cells in the cerebellum (Arin et al. , 1991). These
key GABAergic neurons receive excitatory input from olivo cerebellar climbing
fibers and parallel fiber input from granule cells and , project to the mostly
GABAergic cerebellar nuclei. Despite this important finding of fewer Purkinje cells
057- 064 10/28/03 12:35 PM Page 58
177235-
GENEJ. BLATT
in the autistic brain , there is a paucity of information about the role of the GABAergic
system in autism or whether the GABAergic system is altered in other parts of the
autistic brain. Thus , our recent findings demonstrated statistically significant reduc-
tions in the density of (" H)- flunitrazepam labeled benzodiazepine binding sites and in
H)- muscimollabeled GABA receptors in high binding regions of the hippocampal
formation of autistic individuals (Blatt et al. , 2001). We further found that the alter-
ation in (" H)- flunitrazepam labeled benzodiazepine binding sites showed a trend for
significance for a decrease in the number of binding sites but having normal affinity
to bind the ligand (Blatt et al. , 2002; Guptil et al. , 2003 , submitted). The method and
results of this study are summarized below. Taken together , the results from these
studies suggest a new direction in autism research.
Method
Brain Tissue
Brain tissue was obtained from the Harvard Brain Tissue Resource Center
(HBTRC) in Belmont, Massachusetts. Seven blocks (four autistic and three control)
from fresh , frozen hemispheres stored at - , measuring 10- 5 mm were cut from
the main body of the hippocampus , transported in dry ice to Boston University
School of Medicine , and stored in a - C freezer. The study was conducted blind
using coded case numbers. Detailed medical histories of every autistic case were on
file at the HBTRC. A summary of the cases used in this study appears in Table
Images of the sections and standards were digitized into the autoradiographic
image analysis program , Inquiry (Loats Associates), using a digital camera interface.
The tritium standards were used to construct a calibration table for each film cassette
and the images of all sections exposed in a x- ray cassette were tagged to a calibration
table specific to that cassette. The optical density of specific architectonic regions or
lamina in the hippocampal formation were sampled using the Inquiry system and
converted , via the calibration table , to femtomoles bound per miligram (fmol/mg) of
dry tissue protein.
A total of 11 hippocampal lamina from six hippocampal subfields (dentate gyrus
(DG), CA3 , CA2 , CAl , pro subiculum (Pros), and subiculum (Sub)) were sampled
according to the descriptions of Rosene & Van Hoesen (1987). The eleven lamina
sampled were chosen because they showed either statistically significant differences
or trends toward significance in the single concentration assay, or were high bind-
ing regions. The lamina sampled were: DG: granule cell layer , inner third of the
molecular layer , outer two- thirds of the molecular layer; CA3: pyramidal cell layer;
CA2: pyramidal cell layer; CAl: pyramidal cell layer; Pros: pyramidal cell layer
057- 064 10/28/03 12:35 PM Page 60
177235-
GENEJ. BLATT
molecular layer; Sub: pyramidal cell layer , molecular layer. Plotting the nonspecific
data yielded a linear regression that was subtracted from the total binding measure-
ments (femtomoles/mg tissue) to yield a " specific " binding measurement at each
concentration.
The multiple concentration assay utilized a binding curve ilustrating ligand con-
3 (H)-
centration (nM) vs. fmol/mg of tissue bound. Specific binding of flunitrazepam
to each of the regions studied from each subject was fitted independently with a
hyperbolic binding equation to derive individual estimates of Kd and Bmax for each
region. A two-way nested model , analysis of variance (AN OVA) was used to test sta-
tistical significance. ANOVA is an analysis in which the total variability of the data is
partitioned into compartments attributed to experimental conditions (Freund , 1988).
This analysis is followed up, when statistically significant differences were found
with a posteriori between group comparisons test using the post- hoc Bonferroni test.
Results
stratum
7000
6000
5000
4000
3000
2000
1000
Concentration (nM)
stratum moleculare
7000
6000
5000
4000
3000
2000
1000
Concentration (nM)
3 (H)-
FIGURE 2. Mean flunitrazepam binding ofbenzodiazepine binding sites in stra-
tum moleculare of the subiculum in 4 autistic and 3 control cases across seven con-
centrations of ligand. Note the reduction in the number of binding sites (Bmax) in the
autistic group compared to sex- and age- matched controls. The binding affinity (Kd)
showed no differences.
Results from the multiple concentration saturation binding study indicate that in
ten out of the eleven hippocampal lamina measured , there was a reduction in the
Bmax values in the autistic group compared to the sex- and age- matched controls.
Two graphs in figures 1 and 2 ilustrate the 3 (H)- flunitrazepam binding in two hip-
pocampal laminae (stratum pyramidale and stratum moleculare) in the subiculum
taken from the mean values from the 4 autistic and 3 control cases. The amount of
binding, measured infemtomoleslmg of tissue , is plotted at each of the seven con-
centrations of 3 (H)- flunitrazepam. The binding curve for the autistic group falls below
that of the control group in both figures 1 and 2. Note the separation at the highest
concentration at the right of the curve. Statistical tests showed a trend for significance
for the reduction in Bmax values in the autistic group compared to controls suggest-
ing a change in the number ofbenzodiazepine receptors. The Bmax values represent
on average a 20.1 % reduction in the autistic group for the Will hippocampal lamina
that show a decrease compared to control. The one laminae that shows virtually no
difference is in stratum moleculare of CAl-there is a 2. 9% increase in Bmax value
in the autistic group compared to the control group. The greatest difference is in stra-
057- 064 10/28/03 12:35 PM Page 62
177235-
GENEJ. BLATT
Discussion
Neuropathological abnormalities in the hippocampal formation in autistic indi-
viduals revealed increased cell packing density and a decreased elaboration of the
pyramidal cell dendritic tree in specific CA subfields (Bauman and Kemper , 1985;
Raymond et al. , 1996). In the hippocampal formation , Blatt et al. (2001) conducted
a survey of eight types of neurotransmitter receptors from four different systems
including the serotonergic C(H)- 80H- DPAT labeled 5- receptors and 3 (H)-
ketanserin labeled 5- HT, receptors), cholinergic C(H)- pirenzepine labled Ml recep-
hemicholinium labeled high affinity choline uptake sites), glutamatergic
tors and 3 (H)-
(H)- MK801 labeled NMDA receptors and 3 (H)- kainate labeled kainate receptors)
and GABAergic systems C(H)- flunitrazepam labeled benzodiazepine binding sites
and 3 (H)- muscimollabeled GABA receptors). Data from these single concentration
ligand binding studies indicated that despite the neuroanatomical abnormalities , only
the two GABAergic receptor markers demonstrated a statistically significant reduc-
tion in high binding regions (for 3 (H)- flunitrazepam: stratum pyramidale of CA2 , pro-
subiculum and subiculum and stratum moleculare of the subiculum; for
(H)- muscimol: stratum pyramidale of CAl; Blatt et al. , 2001). A subsequent multi-
ple concentration saturation binding experiment, described here , further demon-
strated a trend for significance that the number of 3 (H)- flunitrazepam labeled
benzodiazepine binding sites were reduced in ten of eleven hippocampal laminae
quantified (decreased Bmax) whereas the binding affinity for the ligand was normal
(Kd) in all eleven laminae (Blatt et al. , 2002; Guptil et al. , 2003 , submitted). These
results suggest that alterations in the GABAergic system are not restricted to the cere-
bellum and that information processing in the hippocampal formation is also altered
in the autistic brain.
An interesting development in autism research with possible implications to the
GABAergic system is the results from the South Carolina Autism Project that have
looked at chromosomal abnormalities in autism. Among the first 100 cases enrolled
in this project, abnormalities of chromosome 15q 11- 13 have emerged as the single
most common finding (about 4% of cases), and candidate genes include genes for
three GABA receptor subunits , GABRa5 , GABR , and GABRy3 (Schroer et al.
1998). Correlating with neuropathology in the hippocampus of autistics , the a5fBy2
subunit combination of GABA receptors predominate on hippocampal pyramidal
cells , making up 20% of the GABA receptor population on these cells (McKernan
and Whiting, 1996; for review of GABA subunit combinations throughout the brain
see Whiting et al. , 2000). An additional subunit combination of GABA receptors has
a widespread distribution in the brain (approximately 50% of the total GABA recep-
tor population) localized to GABA interneurons (al 2y2; Whiting et al. , 2000). It is
j.
unclear whether any of these candidate genes are affected in cases with this chromo-
somal abnormality or in other cases of autism , but it is interesting to speculate , given
the more widespread involvement of the GABAergic system , that subunit gene(s) for
GABA receptors are likely target(s) in autism (e. , subunit switching at specific
times during development possibly due to neuronal loss in specific subcortical or cor-
tical nuclei). Interestingly, Pericak-Vance and colleagues (Shao et al. , 2003) applied
an innovative new statistical genetic approach to families whose children scored high
in the " insistence on sameness " character trait (extreme difficulty with changes to
their daily routine). These researchers discovered a strong link to the GABR )3 gene
on chromosome 15q ll- q 13 in the study group.
Thus , the GABAergic system represents a new directional approach to autism
research. If there are changes in the availability of specific GABA subunit receptors
during critical prenatal and/or early postnatal developmental periods , then the con-
sequences may result in altered GABA receptor function and may seriously impact
the development of afferent- efferent connectivity as well as the delicate excitatory-
inhibitory balance at the synaptic level.
References
Arin , D. , Bauman , M. L. and Kemper , T.L. (1991) The distribution of Purkinje cell loss in
the cerebellum in autism (abstract). Neurol. 41 307.
Bauman , M. L. and Kemper , T.L. (1994) Neuroanatomic observations of the brain in autism
in The Neurobiology of Autism (Bauman , M. L. and Kemper , T.L. , eds), pp. 119- 145.
Johns Hopkins University Press , Baltimore.
Bauman, M. L. and Kemper , T.L. (1985) Histoanatomic observations of the brain in early
infantile autism. Neurol. 35 , 866- 874.
Blatt, Gj., Fitzgerald , C. , Guptil , J.T. , Booker , A.B. , Kemper , T.L. and Bauman , M.
(2001) Density and distribution of hippocampal neurotransmitter receptors in autism: an
auto radiographic study. J. Autism Dev. Disord. 31 , 537- 543.
Blatt, Gj., Bauman , M. L. and Kemper , T.L. (2002) Reduced number ofbenzodiazepine bind-
ing sites and decreased density of GABA receptors in the hippocampal formation of
autistic individuals. Forum of European Neuroscience Abstract 3:484.
Dennis , T. , Dubois , A. , Benavides , J.and Scatton , B. (1988) Distribution of central Wj (benzo-
diazepine ) and w, (benzodiazepine2 ) receptor subtypes in the monkey and human
brain. An autoradiographic study with CH)flunitrazepam and the Wj selective ligand
CH)zolpidem. Pharmacal. Exp. Ther. 247 , 309-322.
Freund , T.F. and Antal , M. 1988. GABA- containing neurons in the septum control inhibitory
interneurons in the hippocampus. Nature 336: 170- 173.
Geary, WA and Wooten , G. F. (1983) Quantitative fim autoradiography of opiate agonist and
antagonist binding in rat brain. j. Pharmacal. Exp. Ther. 225 , 234-240.
Guptil J.T. , Booker , A. , Gibbs , T. , Kemper , T.L. and Bauman M. L and Blatt, Gj. (2003) CH)
flunitrazepam- Iabeled benzodiazepine binding sites in the hippocampal formation in
autism. BioI. Psychiatry (Submitted).
McKernan , R.M. and Whiting, P j. (1996) Which GABA receptor subtypes really occur in the
brain? Trends in Neuroscience 19 139- 143.
Raymond , G. , Bauman , M. L. and Kemper , T.L. (1996) Hippocampus in autism: a Golgi
analysis. Acta Neuropathol. 91 , 117- 119.
057- 064 10/28/03 12:35 PM Page 64
177235-
GENEJ. BLATT
Rosene , D. L. and G. W Van Hoesen (1987) The hippocampal formation of the primate brain:
A review of some comparitive aspects of cytoarchitecture and connections. In Cerebral
Cortex , E. G. Jones and A. Peters (eds. ), Plenum Press , New York , vol. 6 pp. 345- 456.
Schroer , RJ, Phelan , M. , Michaelis , RC. , Crawford , E. , Skinner , S. A. , Cuccaro , M.
Simensen , RJ, Bishop, J., Skinner , C. , Fender , D. , & Stevenson , RE. 1998. Autism and
maternally derived aberrations of chromosome 15q. American Journal of Medical
Genetics 76:327-336.
Shao , Y , Cuccaro , M. , Hauser , E. , Raiford , KL. , Menold , M. , Wolpert, C. , Ravan
, Elston , L. , Decena , K , Donnelly, S. , Abramson , RK , Wright, H. , DeLong,
, Golbert J.R and Pericak-Vance , M. A. (2003) Fine mapping of autistic disorder to
chromosome 15q ll- q 13 by use of phenotyic subtyes. American Journal of Human
Genetics 72: in press.
Shaw , C. , Aoki , C. , Wilkinson , M. , Prusky, G. and Cynader , M. (1987) Benzodiazepine
(CHJflunitrazepam) binding in cat visual cortex: ontogenesis of normal characteristics
and the effects of dark rearing. Dev. Brain Res. , 67- 76.
Whiting, P J, Wafford , KA. , McKernan , RM. , 2000. Pharmacologic Subtyes of GABA
Receptors Based on Subunit Composition , In: GABA in the Nervous System: The View
at Fifty Years. edited by D. L. Martin and RW Olsen. Ch. 8. Pp.113- 126. Philadelphia:
Lippincott Wiliams and Wilkens.
Acknowledgements
Mailing Address:
ABSTRACT. This is the first of two articles offring a new theoretical approach to reading,
based on the Trimodal Model of Brain Organization (Lauter, 7998a 999a b). This general
brain model posits a "neurotypology " of individual difrences created during prenatal develop-
ment and influenced by conditions of exposure to sex hormones, impacting virtually every aspect
of human behavior. As outlined in the current paper, specifc applications of the Trimodal
7) characterization of certain sensorimotor abilities as
Model to reading include:
fundamental
skills for reading; 2) assignment of these skills to the two cerebral hemispheres according to a
new approach to functional asymmetries, the EPIC model; and 3) description of individual dif
ferences in reading ability as depending on difrences in degree of access to right- vs. left- brain
skills. In the second paper (Lauter McKane, this volume), preliminary data testing the the-
ory are provided.
INIODUCTION
The history of dyslexia is notable above all for the consistency and intensity of
controversy surrounding virtually every issue (cf. Duffy & Geschwind , 1985; Hulme
& Snow ling, 1997; Bowan , 2002). Topics of concern have included: 1) how to classify
reading problems; 2) whether sensory processes or only higher- level cognitive func-
tions are important; 3) if there are problems with sensory processing, whether these
are predominantly visual or auditory; and 4) the ultimate cause of reading problems
including the paradox of depressed reading ability in children who are otherwise of
normal intelligence.
Recently we have proposed a new brain model known as the Trimodal Model of
Brain Organization (Lauter , 1998a, 1999a b), which may offer a means of resolving
many of these controversies , in the form of a new biologically-based synthesis which
can accommodate and relate a wide variety of theories and observations. This paper
wil provide an overview of the Trimodal Model , and outline specific predictions
regarding its application to reading and reading problems. The second paper (Lauter
& McKane , this volume) wil summarize some preliminary data collected to test these
predictions.
065- 084 10/16/03 5 : 38 Page 66
177235-
JUDITH L. LAUTER
Second , with regard to complementary functions of the two sides of the brain , the
EPIC model considers the right hemisphere to be responsible for processing in
extrapersonal space (for example , features of sounds and visual patterns which can per-
ceived at a distance from the body surface) while the left hemisphere specializes in
tasks performed in peripersonal space (such as sounds and visual patterns which must
be very close to the ear and eye to be perceived). This distinction is similar to one
suggested by Heilman (1994), but the EPIC model extends this to posit analogues in
all four main somatic functional systems- auditory, visual , somatosensory and motor
(Table II). The idea is that in most everyday tasks , the two sides of the brain , with
their complementary abilities in extra- vs. peri- personal space , work together
accomplish tasks which neither could do alone (for a more complete discussion , d.
Lauter , 1999b). Laboratory research on asymmetries , which may focus on one side
to the exclusion of the other , can lose sight of the bilateral nature of most behaviors.
Third , as implied in Table the EPIC model does not consider these righUleft
specializations in terms of abstractions (such as " holistic vs. analytic ), but as physio-
logical cortical extensions of biophysical properties which are known to distinguish
classes of sensory receptors and motor effectors at the body periphery (Lauter , 1987
1997 , 1999b). While there are currently no data documenting anatomical distinctions
in connectivity between , for example , Pacinian corpuscles in the skin and the right
vs. the left hemisphere , there are a variety of known mechanisms for top-down selec-
tivity, which could be ultimately reflected in functional (and physiological) differ-
ences in the coding of cell populations at the level of the cortex. It should be possible
to document such physiological distinctions in humans by means of noninvasive
brain imaging, as long as analysis focuses on individual brains , since averaging across
subjects can readily obscure what can be very dramatic individual differences in
asymmetric patterns of activation (cf. Lauter , 1982 , 1983 , 1984 , 1992; Lauter et al.
1985 , 1988).
formance distinction: it assumes that all brains are born with all the tools in place
(and assigned in the same way to the two hemispheres), but not all brains have ready
access to all the tools.
B. The Trimodal Model: Individual differences. According to the model , these
sorts of individual differences are established during prenatal development under the
same type of hormonal influences (primarily related to testosterone exposure) which
others have concluded are important for righUleft growth patterns (reviewed in
Nyborg, 1994). The Trimodal Model posits that a continuum of prenatal testosterone
exposure (from zero to high) results in a continuum of right/left configurations character-
ized by a continuum of "salience of right- vs. left side specializations. For several bio-
logical and ethological reasons (cf. Lauter , 1998a , 1999b), the continuum is predicted
Table II. Proposed dimensional and receptor/effector bases for hemispheric specializations
(EPIC model of functional asymmetres)
EXTRAPERSONAL SPACE PERIPERSONAL SPAC EVERYDAY EXAMPLES
(Right hemisPhere) (Left hemisPhere) (2 examples/set)
Given the " unequal" responsibilities of the right vs. left hemispheres mentioned
above , plus the posited " disconnection " actions of prenatal testosterone (cf. Lauter
1998a , 1999b), it can be concluded that not all of these brains wil have equal access
to all skils. However , the Trimodal Model suggests that all these brains should be
considered to be normal from a biological point of view , since all would be equally
adaptive for the types of societies in which mammalian brains developed for milions
of years. The underlying features of neural function which differ somewhat from
brain to brain are expressed in a wide variety of individual characteristics , ranging
from personality to constellations of skils (such as those required for reading) to sus-
ceptibility to a variety of clinical conditions.
First , the Trimodal Model predicts that the brain which develops in a testos-
terone- free environment (thus predictably most typical of women) is the one in which
the right hemisphere comes to be the more salient. (This may be indexed by the
larger size of certain regions surrounding the right- side Sylvian fissure , areas which
have been the focus of much recent research on anatomical asymmetries and learn-
ing disorders- d. Geschwind & Galaburda , 1987; Galaburda , 1994; Lewis &
Diamond , 1994; Hynd & Hiemenz , 1997). Due to the " polypotent" nature of the right
hemisphere , these brains should have access to all skils-both those which are spe-
cializations of the right brain (processing in extrapersonal space , management of gen-
eral health , neural coordination , etc. plus left- brain specialities for processing in
peripersonal space which are overseen and coordinated by the right side (cf. the skils
profiles in the middle panel of Fig. 1). Thus the Trimodal Model refers to these brains
as " polytropic " (many- skiled), or as " right brain/whole brain " types.
With increasing testosterone exposure , this default pattern is gradually under-
mined , a state which may interfere with access to the skils of either side. As the rel-
ative salience of the two sides approaches a condition of symmetry, both categories
of functional specializations may become relatively inaccessible , retaining only excel-
lence for righUleft coordination of gross motor control , such as required for athletics.
The Trimodal Model refers to these as " middle " brains , both because of their devel-
opment under mid- range levels of testosterone exposure , and because of their sym-
metric anatomical and physiological properties.
At increasingly higher testosterone levels , the left hemisphere may become over-
whelmingly salient and actually depress growth of the right side. This can create " left-
brain " individuals who , while they may excel in one or more skils provided by the
left side , such as fine motor control or phonemic awareness , at the same time may
also suffer from diminished right-brain participation , expressed in different cate-
gories of il health , poor social skils , and a lack of motor coordination.
!,.
JUDITH L. LAUTER
PhohBrnic awarenes
Pa-vo vi9.aI
htlagno \Asual
2- eye coordif1ation
Adltal ;rn q;ery
(1) Phonemic awareness. One fundamental skil required as a foundation for read-
ing is the auditory ability known as phonemic awareness , which is defined within the
EPIC and Trimodal models as the capacity to distinguish and separate the acoustic
cues for phonemes when embedded within syllables. (As used here , the term does not
refer to the ability to identify phonemes in isolation , or to learn to map them onto
alphabetic symbols , which virtually all brain types can do. ) Acoustic cues for
phonemes are definitely associated with processing in since they are
peripersonal space
extremely difficult to discern if a listener is very far removed from the speaker.
Current estimates suggest that approximately one- third (30- 40%) of the general
population exhibits a lack of this skil (evidence reviewed in Lindamood et al. , 1997).
Since a lack of phonemic awareness may impact on other skils , such as the ability to
understand speech in a noisy background (the trademark of " central auditory pro-
cessing disorder ), auditory attention , information retrieval using the auditory chan-
nel (via spoken or recorded instructions or lectures), and the acquisition of
phoneme- level grammatical markers , many members of this relatively large segment
of the population may present with a variety of other troubles in addition to reading,
which compounds their problems in school (Lauter et al. , 1998; Lauter , 2000b).
(2) Visual acuity. A second left-brain skil necessary for reading, particularly given
the font sizes which begin to be used toward the end of the elementary grades , is
visual acuity. In this context , visual acuity refers not only to performance of the
peripheral visual mechanism (cornea through retina), but also to posited specializa-
tions of the left hemisphere for processing small , static visual patterns captured on the
foveal portion of the retina in peripersonal space. (As indicated in Table II , this
implicates the " parvocellular" component of the visual system , which has been iden-
tified as originating in foveal retinal cells , and extends through thalamus into the cor-
tex. ) The relatively high ilumination typically used for reading represents an
additional factor demanding left-brain foveal processing, since the retinal cones
which predominate at the fovea , respond well at high light intensity while the
Table III. Skills related to readng & spellng provided by each hemisphere
RIGHT HEMISPHERE LEFT HEMISPHERE
Visual word shapes & lengths letter shapes (especially small font sizes)
mental imagery visual function in bright light
Auditory syllable & sentence " rhythms phonemic awareness (within- syllable acoustic cues)
Somatosensory vibration of larynx for voiced sounds local sensation for articulation targets (tongue tip, etc.)
Motor jaw open/close of syllabic " rhythm fine- motor kinesthetic feedback from speech articulation
eye- movement coordination
065- 084 10/16/03 5 : 38 Page 73
177235-
response of rods (which predominate in extrafoveal retina) falls off with higher lev-
els of ilumination.
Experimental data regarding visual acuity of the " cortical" type , that is , individ-
ual differences over and above those which can be accounted for by differences in
peripheral function alone , are represented primarily by studies on functional asym-
metries related to visual spatial frequency (cf. Sergent 1983). Although some authors
have concluded that research on processing asymmetries for spatial frequency is
inconclusive (cf. Sergent , 1994), in general these findings do suggest that higher spa-
tial frequencies (such as text printed in font sizes 12 pt or less) may be preferentially
processed by the left brain.
It is notable that the font sizes used in elementary- school reading materials
decrease dramatically from kindergarten to third grade. This suggests that some of
the reading problems which " emerge " around third grade may be the expression of
brains which simply do not have sufficient access to left-brain acuity to accomplish
accurate identification of letters printed in 12- point type , particularly when presented
in continuous text where identification must be very rapid. In fact, reading can be
improved in some children simply by allowing them to read text printed in larger
type (Cornelissen et al. , 1991). (It is possible that such problems may be exacerbated
by additional difficulties with eye movements, discussed in the next section.
(1) Mental imagery. The importance of mental or " concept" imagery for reading
comprehension has been cited by many authors over the years (cf. Lindamood et al.
1997 for a review). For instance , these latter authors note that an excellent predictor
of abilities for mental imagery and accomplished reading comprehension is a test of
design reproduction , where an individual is required to hold the features of a geo-
metric figure in mind and then redraw it from memory. Being able to create , retain
and recall mental images is clearly related to abilities for identifying visual objects at
a distance , which we have previously suggested constitute a right-brain specialty
(Lauter , 1997 , 1999b), since both types of task emphasize overall " gestalt" form based
on details comprehended in an overall form , and object identification involving
movement instead of static properties.
Ready access to mental imagery during reading may also be related to an indi-
vidual's ability to remember the images of dreams. Some reading experts have noted
that children whose reading comprehension is impaired by poor mental imagery
often do not " remember " their dreams , and one of the most dramatic side- effects of
training in imagery is that these individuals begin being able to report and describe
dreams in great detail , starting with black and white images and progressing to color
(Bell , personal communication). It has been known for some time that dreaming
occurs primarily during rapid eye movement (REM) sleep (Foulkes , 1966; see next
section) with higher activity over the right as opposed to the left hemisphere
(Asenbaum et al. , 1995 , Benca et al. , 1999), findings that also seem to support
065- 084 10/16/03 5 : 38 Page 74
177235-
JUDITH L. LAUTER
fovea of the retina. The parvocellular part of the visual system mentioned above is
employed during the fixations to capture the small , static visual image of a few let-
ters on the foveal retina.
However , if only the parvo system were available , reading rates would have to
be extremely slow , since cells of this system are " slow- adapting, " that is , they take a
long time to " lose " one image when confronted with a new one. If a reader moved
the eyes quickly to a new image (a new string of letters), and had only the parvo sys-
tem active , the impression would be of blurring, as the parvo system s trace of the
old image gradually disappears and the new one is registered. (In some children with
reading problems , this is exactly the percept reported , which has led to one theory
of dyslexia , discussed below.
However , there is another component of the visual system which seems to help
out when a fairly rapid sequential series of fixations is required. This is the " magno-
cellular " system , where cells are " rapid- adapting, " that is , they have the property of
being able to start and stop responding quite quickly. It has been suggested (ef.
Breitmeyer & Ganz , 1976) that during fluent reading, the two systems work in tan-
dem- that is , the parvo system is active during the static fixations , working to capture
the image for transmission to the brain , while the magno system takes over during
the subsequent saccade , working to " erase " the remains of the previous image in
preparation for receiving the next one.
A number of properties of the magnocellular visual system make it a candidate
for right- brain specialization according to the Trimodal Brain model: 1) it seems to
be very important for eye movements (cf. Eden et al. , 1995), which as we have said
are clearly adaptive for extrapersonal space perception (cf. Stein , 1989); 2) the magno
system arises from cells having fairly large receptive fields away from the fovea , thus
,"
065- 084 10/16/03 5 : 38 Page 75
177235-
supporting wide- scan monitoring along with gestalt whole- form " perception; and 3)
those cells depend primarily on rods , and are thus more sensitive to light/dark con-
trasts than to color , and are also active at low luminance , all very adaptive features
for vision at a distance.
The fact that extrafoveal rods are much more reactive than foveal cones both at
low luminance levels and in response to blue light (ef. the " blue-blind" fovea-Wilmer
1946) may account for the observation that atypical eye movements can sometimes
be improved by the use of gray or blue filters (cf. Wiliams et al. , 1992; Solan et al.
1998), which presumably work by recruiting or " waking up " a relatively inactive
magno system. (By the same logic , it is not surprising that red filters do NOT improve
eye movements (cf. Wiliams et al. , 1992), since it is the parvo- system cells of the
fovea , not the extrafoveal magno- system cells , which are stimulated by red.
This and other signs of difficulties attributable to reduced magnocellular function
have led several researchers to posit that some reading problems are due to a " tran-
sient or magnocellular deficit " suggested as an actual neurological abnormality
affecting this component of the visual pathways (cf. Stein & Walsh , 1997; Solan et al.
1998). In contrast , the Trimodal Model attributes such problems to a more dynamic-
and thus trainable- lack of access " to right-brain specializations , as wil be discussed
more below.
(3) Eye- movement coordination. The large body ofresearch on eye movements dur-
ing reading also provides overwhelming evidence of the importance of coordinated
functions in eye movements (Garzia & Sesma , 1993; Garzia & Peck , 1994). There are
two categories of coordination, or synchronization , which are predictably overseen
by the right brain.
(3- a) Synchronization of right and left eyes. First , for accurate integration of images
captured by the right and left eyes , it is clearly necessary to ensure that both eyes
move together , particularly in a demanding visual task such as reading. This type of
synchronization is not an obligatory result of hard wiring- for instance , it shows a
clear developmental time course (the right and left eyes of neonates very often move
independently, though even in neonates they become more coordinated under low-
luminance conditions , again implicating the magno system- Haith & Goodman
1982), and righUleft coordination requires active , waking " high- level" control (right
and left eyes are not coordinated during REM sleep- Zhou & King, 1997).
Other authors , drawing on data from normal as well as brain- injured individuals
have concluded that stereoscopic vision , accomplished by the vergence system
depends on the right hemisphere (cf. comments in Eden et al. , 1995). The Trimodal
Model agrees with this , and extends such a conclusion to include all aspects of two-
eye coordination , as representing just another instance of bilateral motor coordina-
tion (analogous to coordinating right and left hands in bimanual tasks)- thus the
model classifies two- eye coordination in general as a specialty of the right hemi-
sphere. Such an assignment is additionally supported by research on individuals with
unilateral brain damage , where there are indications that eye movements are dis-
rupted more , and in different ways , following right- side as opposed to left- side injury
(Weinberg et al. , 1979 , Gordon et al. , 1985; Zihl , 1995).
065- 084 10/16/03 5 : 38 Page 76
177235-
JUDITH L. LAUTER
Given these very specific assignments of different skils and capabilities to the two
cerebral hemispheres , it is easy to generate profiles of different possible subtypes of
reading problems , which should not only account for all of the types previously
observed , but also be highly predictable regarding how different combinations of
problems may characterize specific individuals. Such a systematic " bottom-
approach may lead to more objective and biologically- based guidelines for classify-
ing and categorizing the different types of reading problems , which may prove use-
ful for addressing clusters of features , designing and selecting remediation , and
documenting treatment efficacy (cf. comments regarding the need for " appropriate
tests " to address different populations in Stein & Walsh , 1997).
The Trimodal Model's suggestion that skil profiles arise from a trimodal distri-
bution of brain types also suggests the biological origins of the kinds of numbers that
have been associated with the incidence of reading problems- such as 20- 30% of pub-
lic- school populations having severe difficulties with reading (Shankweiler &
Liberman , 1989), 48% of the adult population being without functional literacy skils
(National Center for Education Statistics , 1993), and 60% of the adult prison popula-
, "
tion lacking everyday literacy skils (Conway, 1993). These numbers can seem over-
whelmingly large if we continue to think of reading difficulties as signs of neurologi-
cal abnormality-but , in fact, the size of these incidence statistics suggests rather that
these are signs of normal variation just as predicted by the Trimodal Model (Lauter
2000a, 2001 , 2002). As we wil see , the Trimodal Brain approach to reading also helps
us come a little closer to grasping the origins of the different reading profiles together
with their co- morbidities , which lacking such a context can be difficult to understand.
The possible subtypes of reading problems suggested by this outline include dif-
ficulties arising from a lack of anyone of the skils listed above , or the lack of difrent
combinations of skils- two , three , four , etc. all of which should result in very different
behavioral profiles , pointing to very different patterns of access to cerebral special-
izations.
By way of ilustration , consider again the neurological categories from the top
panels of Fig. 1 , together with the notations regarding the right vs. left- brain skils we
have just discussed , summarized in the bottom panel.
(1) Right- brain/Whole- brain individuals (right side of Fig. 1). These individuals are
considered to be those who develop in a testosterone- free intrauterine environment.
They are characterized as having excellent access to the specializations of both sides
of the brain , and are therefore predicted to have no difficulties at all in reading. These
are the " natural readers " who " catch on quickly " when first shown the correlation
between phonemes and alphabetic symbols , who " see pictures in their head" when
they read , naturally visualize words to be spelled , and continue to enjoy reading as a
lifelong entertainment.
Their phonemic awareness wil follow or progress ahead of the developmental
course predicted by current norms (such as those established for the Lindamood
Auditory Conceptualization (LAC) Test- cf. Lindamood et al. , 1992 , 1997), and thus
achieve an adult level for nonsense sequences of four phonemes before or no later
than 7th grade. They wil exhibit vivid and accurate mental imagery, and their eye-
movement coordination , in terms of both two- eye synchronization and rhythmic eye
tracking, should progress along the expected developmental course (for instance
according to norms collected with the Visagraph II system- d. Colby et al. , 1998).
(2) Left- brain individuals (left side of graph). The Trimodal Model suggests that
these individuals , who are assumed to develop under the highest levels of exposure
to prenatal testosterone , wil be limited primarily to left-brain skils such as phone-
mic awareness-which , however , they may show in a very advanced hyperlexic
form. (This may be seen in children with left- brain conditions such as Asperger
Syndrome- Lauter et al. , Submitted). However , their diminished access to right- brain
skils should result in the disastrous combination of poor to absent mental imagery
and extremely irregular eye movements , which wil not develop in a normal way and
wil continue to impede reading performance so that it remains well below that pre-
dicted for their age.
Note that their abnormal eye movements should include a lack of coordination
of the two eyes plus signs of irregular tracking-both interpreted as primary signs of a
lack of access to right-brain abilities. (Since parvo-visual function should be intact in
065- 084 10/16/03 5 : 38 Page 78
177235-
JUDITH L. LAUTER
these individuals , their eye- tracking features may show inconsistent abnormalities
very short fixation time combined with very many regressions , etc. ) This profile of
eye- movement problems should distinguish them from certain middle- brain types
(see next section), in whom the two eyes are well synchronized , but eye tracking is
disturbed in a consistent way (long fixation times and multiple regressions , etc. ), thus
interpreted as a secondary result of poor access to left- brain skils for parvo- system
function. The left-brain types should be those who are most helped by the use of gray
or blue filters , which predictably act by " turning up " the magno visual system , which
may be said to be " sleeping " (due to the diminished access to right-brain participa-
tion) in this type of brain.
Some authors have reported that eye movements can be improved through the
use of blue filters- in some cases , improving performance in as many as 70- 80% of
reading- disordered subjects (cf. Lovegrove , 1993; Demb et al. , 1998; Iovino et al.
1998; Solan et al. , 1998; Robinson & Foreman , 1999). Such a high proportion may
be accounted for by the fact that, in some studies (such as Solan et al. , 1998) children
with " significant decoding problems " (i. , poor phonemic awareness , or what we
would predict are " poor- parvo " children) were screened out of the study, thus creating
a study population which should have a very high proportion of " poor- magno " (i.
blue- responding) children.
Left- brain individuals may be able to obtain meaning from text by rote memo-
rization of word definitions and laborious re- reading, and may depend heavily on
assistance from ilustrations and other extra- text cues to meaning, which they cannot
obtain for themselves given their lack of mental imagery. These characteristics of
their reading performance should appear in combination with other features which
the Trimodal Model associates with the left- brain type , such as strong right- handed-
ness for fine motor tasks , poor ' motor planning ' skils , diminished social skils and
empathy, inappropriate or monotonal speech " melody, " and a primitive sense of
humor (these are the children who often " don t get the joke
They wil also be characterized by " released" brainstem activity combined with
poor skils for maintaining attention , leading to versions of attention deficit, sensory
hypersensitivity, and/or motor hyperactivity (cf. Lauter , 1998b , 1999a b)- thus these
may represent the " dyslexia- plus " children described by Denckla (1985) and others.
They may also show depressed immune function , with associated conditions such as
chronic otitis media and allergies; and exhibit signs of poorly managed internal body
functions , such as erratic sleep cycles , poor eating habits , frequent gastrointestinal
upset, and headaches.
(3) Middle- brain individuals (middle panel of Fig. 1). These individuals , exposed to
moderate amounts of prenatal testosterone (whether they are XY or XX) should
together represent a continuum of diminished access to skils of both sides. For exam-
ple , those on the right side of middle may retain quite good access to right-brain skils
such as imagery and two- eye synchronization , but lack phonemic awareness. Eye-
tracking movements during reading may be perturbed secondary to their impaired
access to parvo-visual functions- thus in these individuals , eye- movement abnormal-
065- 084 10/16/03 5 : 38 Page 79
177235-
ity may be the result of diminished activity in the parvo system , and not a cause of
reading problems in and of itself.
Their reduced (parvo- system) ability for rapid recognition of letters during visual
fixation may lead to eye tracks with a high number of fixations (" one fixation per let-
ter ), long fixation times (" what IS that letter?" ), and multiple regressions (" going
back to look again ). Note that such features of eye tracking predicted for these chil-
dren (considered neurologically normal according to the Trimodal Model) resemble
those observed in adults subsequent to left-brain damage (e. , Zihl , 1995).
It may even be possible that as these types of children move through the early
school grades , their lack of phonemic awareness conflicting with classroom demands
may result in an actual degradation of eye movements. This could account for the
observation that by third grade , the incidence of children with perturbed eye move-
ments is often higher than that measured in children two years younger (McKane
unpublished data). In general , these " right- middle " individuals may exhibit the type
of reading difficulty referred to as " dysphonetic " with phonemic awareness the over-
riding lack.
Due to diminished access to left-brain skils in general , reduced phonemic aware-
ness wil predictably be accompanied by other signs of reduced left-brain access
such as ambiguous hand dominance for fine- motor control tasks (thus these individ-
uals may be ambidextrous to mildly left- handed) or generally impaired fine motor
control. A lack of fine- motor sensitivity may also result in a lack of awareness of the
position of the speech articulators (resulting from poor feedback from muscle and
joint receptors), as has been observed in dysphonetic children (cf. comments by
Heilman et al. , 1996 and Lindamood et al. , 1997). An additional failure based on lack
of access to left- brain abilities , in this case for fine somatosensory perception (cf.
Table II), may result in such individuals having a poor sense of touch on the skin
around the mouth (such as from particles of food), as has also been noted (ef.
Heilman et al. , 1996).
On the other hand , individuals more in the middle ofmiddle might have dimin-
ished access to the specializations of both sides , resulting in a plethora of difficulties
with reading. For these individuals (perhaps those sometimes called " dysphoneidet-
ics cf. Stein & Walsh , 1997 , or " dyslexia pure Denckla , 1985), both phonemic
awareness and all aspects of eye movements may be well below grade level (cf. a sub-
set of children studied by Eden et al. , 1994), and there may be little or no experience
of mental imagery. They should be more or less ambidextrous (with a variety of types
of " left- handedness ), and may actually excel at whole-body movements involving
coordination of both sides of the body, but should have relatively poor fine motor
control of any body part, including either hand. They should exhibit relatively low
susceptibility to " left- brain ils " such as hyperactivity and psychiatric or immune- sys-
tem problems , but may be fairly vulnerable to weight gain and associated conditions
such as heart problems , though these wil be of somewhat different types than those
experienced by left brains.
Finally, individuals to the left of middle should have even more diminished access
to right- side skils , combined with a " warming up " of left-brain skils. Their phone-
065- 084 10/16/03 5 : 38 Page 80
177235-
JUDITH L. LAUTER
mic awareness may not be as limited as is true for other types of " middle brains " but
their eye movements should resemble the pattern we have suggested wil pertain in
left brains , that is a combination of poor two- eye synchronization plus irregular eye
track patterns which cannot be accounted for by a consistent problem with parvo-
visual function (this may be the condition sometimes referred to as " dyseidesia
As a result of the incomplete nature of their access to left-brain skils , these peo-
ple may also be those who exhibit developmental stuttering which persists beyond
childhood (eye- movement problems have been seen in some stutterers- d. Brutten et
al. , 1984), and they may also be prone to develop certain types of psychiatric prob-
lems such as schizophrenia and depression. It is notable that there is a considerable
literature on perturbed eye movements in schizophrenia (reviewed in Zahn , 1986),
which is controversial in that these problems do not occur in all schizophrenics.
However , this is not a surprising outcome if this psychiatric condition in fact is rep-
resented , as predicted in the Trimodal Model , along a restricted range of the brain-
type continuum (perhaps bridging between left- middle and left), with associated
variations in accuracy of eye- movement control.
SUMMAY
In this paper we have suggested that a new approach to understanding individ-
ual differences in access to right- vs. left-brain skils , the Trimodal Model of Brain
Organization , offers new insights into the biological origins of reading problems. In
addition , the " neurotypology " outlined by the Model also gives us a basis for pre-
dicting which children wil have which types of reading difficulties-whether none at
all (polytropic- brain types), different combinations of problems (middle- brain types),
or skils limited to those supported by the left hemisphere only (left- brain types).
Based on the trimodal distribution of these brain types posited by the model , the gen-
eral conclusion is that reading may in fact be " unnatural" for approximately 2/3 of
the population- adults as well as children (Lauter 2000a, 2001 , 2002). Although this
may seem daunting, there is also a hopeful side to this picture- namely, the posited
neurobiological bases of these difficulties , that is , that they are problems of access
rather than " missing modules. " As a result , intensive neurorehabilitation training
programs specifically targeting undeveloped skils can be used to " wake up " access
to the skils , and add them to an individual's repertoire. Such possibilities wil be dis-
cussed briefly in the second paper (Lauter & McKane , this volume).
As a theory, the Trimodal Model may offer a useful means of bringing together
and triangulating details from many areas of dyslexia research. At the same time
from an empirical standpoint, the Trimodal Model is descriptive , highly predictive
and eminently testable. The companion paper (Lauter & McKane , this volume) pres-
ents preliminary data testing one prediction regarding the co- occurrence of a pair of
fundamental abilities needed for reading-the left- brain skil of phonemic awareness
(PA) and eye- movement coordination (EMC), a right-brain specialty. Specifically, the
prediction is that four general populations of children can be defined , representing
), ),
REFERENCES
Asenbaum , S. , Zeithofer , J., Saletu , B. , Frey, R , Brucke , T. , Podreka , I. Deecke , L. (1995).
Technetium- 99m- HMPAO SPECT imaging of cerebral blood flow during REM sleep
in narcoleptics. Journal of Nuclear Medicine , 1150- 1155.
Benca , RM. , Obermeyer , W. , Larson , C. , Yun , B. , Dolski , I. , Kleist, KD. , Weber , S.
Davidson , RJ. (1999). EEG alpha power and alpha power asymmetry in sleep and
wakefulness. Psychophysiology, , 430-436.
Bowan , M. D. (2002). Learning disabilities , dyslexia , and vision: a subject review. Optometry, 73
553- 75.
Breitmeyer , B. G. & Ganz , L. (1976). Implications of sustained and transient channels for theo-
ries of visual pattern masking, saccadic suppression , and information processing.
Psychology Review , 1- 36.
Brutten , GJ., Bakker , K Janssen , P. , van der Meulen , S. (1984). Eye movements of stuttering
and nonstuttering children during silent reading. Journal of Speech and Hearing Research
, 562- 566.
Laukkanen , H. , Yolton , RL. (1998). Use of the Taylor Visagraph II system to
Colby, D. ,
evaluate eye movements made during reading. Journal of the American Optometric
Association , 22- 32.
Conway, J. (1993). A solution to the iliteracy crisis. Missouri Reader 28-30.
Cornelissen , P., Bradley, L. , Fowler , S., Stein J. (1991). What children see affects how they
read. Developmental Medicine and Child Neurology, , 755- 762.
Demb J.B. , Boynton , G. , Best, M. , et al. (1998). Psychophysical evidence for a magnocel-
lular deficit in dyslexia. Vision Research , 1555- 1559.
Denckla , M. B. (1985). Motor coordination in dyslexic children: Theoretical and clinical impli-
cations. In F.H. Duffy & N. Geschwind (Eds. Dyslexia: A neuroscientifc approach to clini-
cal evaluation (pp. 187- 195). Boston: Little , Brown & Co.
Duffy, F.H. & Geschwind , N. (Eds. ) (1985). Dyslexia: A neuroscientifc approach to clinical evalua-
tion. Boston: Little , Brown & Co.
Eden , G. , Stein , J.F. , Wood , H. , Wood , F.B. (1994). Differences in eye movements and
reading problems in dyslexic and normal children. Vision Research , 1345- 1358.
Eden , G. , Stein J.F. , Wood , M. , Wood , F.B. (1995). Verbal and visual problems in reading
disability. Journal of Learning Disabilities , 272- 290.
Flowers , D. L. (1993). Brain basis for dyslexia: A summary of work in progress. Journal of
Learning Disabilities , 575- 582.
Foulkes , D. (1966). The psychology of sleep. NY: Scribners.
Galaburda , A.M. (1994). Anatomic basis of cerebral dominance. In RJ. Davidson & K
Hugdahl (Eds. Brain asymmetry (pp. 51- 73). Cambridge MA: MIT Press.
Garzia , RP. & Sesma, M. A. (1993). Vision and reading. I. Neuroanatomy and electrophysiol-
ogy. Journal of Optometry and Visual Development , 4- 51.
Garzia , RP. & Peck , C. (1994). Vision and reading. II. Eye movements. Journal of Optometry
and Visual Development , 4- 37.
), ),
JUDITH L. LAUTER
Lauter j.L. (2001). Is reading ' unnatural' for 2/3 of our children? - Right brain , left brain , and
the " missing link" skils for reading and learning. Invited keynote presentation to
SouthWest and Rocky Mountain division of American Association for the Advancement
of Science , Denton TX.
Lauter , j.L. (2002). Is reading ' unnatural' for 2/3 of children and adults? Presented to
American Speech- Language- Hearing Association , Atlanta GA
Lauter , j.L. , Herscovitch , P. , Formby, C. , Raichle , M. E. (1985). Tonotopic organization in
human auditory cortex revealed by positron emission tomography. Hearing Research, 20
199-205.
Lauter j.L. , Herscovitch , P. , Raichle , M. E. (1988). Human auditory physiology studied with
positron emission tomography. Inj. Syka and RB. Masterton (Eds. Auditory Pathway
(pp. 313- 317). NY: Plenum.
Lauter , j.L. , Richey, H. , Gilmore , S., Lynch , O. (1998). Putting the " central" back in central
auditory processing. Journal of Developmental and Learning Disorders , 51- 106.
Lauter , j.L. , Richey, H. , Giddens , C. Culbertson (Submitted). Neurological profie of
Asperger Syndrome: A case study.
Lewis , D. W & Diamond , M. C. (1994). The influence of gonadal steroids on the asymmetry of
the cerebral cortex. In RJ. Davidson & K. Hugdahl (Eds. Brain asymmetry (pp. 31-50).
Cambridge , MA: MIT Press.
Lindamood , P. , Bell , N. , Lindamood , P. (1992). Issues in phonological awareness assessment.
Annals of Dyslexia , 242-259.
Lindamood , P. , Bell , N. , Lindamood , P. (1997). Sensory- cognitive factors in the controversy
over reading instruction. Journal of Developmental and Learning Disorders , 143- 182.
Lovegrove , W. (1993). Do dyslexics have a visual deficit? In S. F. Wright & R Groner (Eds.
Facets of dyslexia and its remediation (pp. 33- 49). Amsterdam: Elsevier Science
Publishers.
MacNeilage , P. F. (1987). The evolution of handedness in primates. In D. Ottoson (Ed. Duality
and unity of the brain (pp. 100- 113). NY: Plenum.
Maples , W. C. (2003). Visual factors that significantly impact academic performance. Optometry,
, 35- 49.
National Center for Education Statistics (1993). National Adult Literacy Survey. Washington DC:
NCES.
Nyborg, H. (1994). Hormones, sex and society. Westport CN: Praeger.
Robinson , G. L. and Foreman , PJ. (1999). Scotopic sensitivity/Irlen syndrome and the use of
coloured filters: a long- term placebo controlled and masked study of reading achieve-
ment and perception of ability. Perceptual and Motor Skills , 83- 113.
Sergent j. (1983). The role of the input in visual hemispheric asymmetries. Psychology Bulletin
481-512.
Sergent, j. (1994). Hemispheric contribution to face processing: Patterns of convergence and
divergence. In RJ. Davidson & K. Hugdahl (Eds. Brain asymmetry (pp. 157- 181).
Cambridge , MA: MIT Press.
Shankweiler , D. & Liberman , LY. (1989). Phonology and reading disability. Ann Arbor: Univof
Michigan Press.
Solan , H. , Ficarra , A , Brannan j.R , Rucker , F. (1998). Eye movement effciency in normal
and reading disabled elementary school children: Effects of varying luminance and
wavelength. Journal of the American Optometric Association , 455-464.
Stein j.F. (1989). Representation of egocentric space in the posterior parietal cortex.
QJarterly
Journal of Experimental Physiology, , 583- 606.
Stein , j.L. & Walsh , V. (1997). To see but not to read: The magnocellular theory of dyslexia.
Trends in Neuroscience , 147- 152.
Weinberg, j., Diler , L. , Gordon , W. , Gerstman , LJ., Lieberman , A. , Lakin , P. , Hodges , G.
Ezrachi , O. (1979). Training sensory organization in people with right brain damage.
Archives of Physical Medicine and Rehabilitation , 491-496.
), ), (pp.
JUDITH L. LAUTER
Mailing Address:
An audio- visual overview of some of these issues (Trimodal Brain model , the AXS
Test Battery, applications to reading) is provided in: J.L. Lauter (2002) Neuroimaging:
How Understanding Individual Difrences Can Improve Your Clinical Practice. (3- hr edu-
cational video with manual) Rockvile , MD: American Speech- Language- Hearing
Association. Available from ASHA professional catalog (1- 888- 496- 6699), item
0112397.
77235- 085- 096 10/17/03 11:09 AM Page 85
ABSTRACT. This paper reports the results of an exploratory study designed to test some of
the predictions of the Trimodal Model of Brain Organization (Lauter, 7998a, 7999a , this
volume). This general brain model posits a "neurotypology " of individual difrences created
during prenatal development and influenced by conditions of exposure to sex hormones, impact-
ing virtually every aspect of human behavior. As outlined in the fturth article referenced above
(Lauter, this volume), applications of the Trimodal Model to reading include: 7) characteriza-
tion of certain sensorimotor abilities as fundamental skills ftr reading; 2) assignment of these
skills to the two cerebral hemispheres according to a new approach to functional asymmetries
the EPIC model; and 3) description of individual difrences in reading ability as depending
on difrences in degree of access to right- vs. left- brain skills. In this article, preliminary data
testing the theory are provided.
INIODUCTION
The synthesis offered in the previous paper (Lauter , this volume) provides a
means of resolving a number of current theories and observations about reading and
reading problems. However , the predictions of that model remain to be tested empir-
ically. To some extent, this can be done through reviews of previous data , some of
which were mentioned in that paper. Additionally, the approach generates a great
many highly testable predictions , particularly regarding the clustering of skils , which
promises to lead to a more biologically- based classification of the variety of reading
problems.
We have begun testing a few of these predictions. For instance , preliminary data
documenting positive correlations between performance on a test of phonemic aware-
ness (the Lindamood Auditory Conceptualization (LAC) Test- cf. Lindamood et al.
1997) and a measure of
fine motor control- both predicted by the Trimodal Model to
be left-brain skils-have been reported (Lauter , 1999a). Testing has also provided
085- 096 10/17/03 11:09 AM Page 86
177235-
initial confirmation (Lauter , 1998a , 1999a) of the prediction that good phonemic
awareness (as assessed with the LAC) is correlated with a condition of hemispheric phys-
iological (qEEG) asymmetry-whichever side is favored (Lauter , 1999a).
In this article we report for the first time test results in children who were exam-
ined both for phonemic awareness using the LAC Test , and for eye movements with the
Visagraph II (Taylor Associates; cf. Colby et al. , 1998). These test instruments were
selected in the interest of employing tools which are reliable , show good subject
acceptance , have established norms , and are also sufficiently inexpensive to encour-
age replication of these preliminary data as well as practical applications in clinics
and schools.
Design and goals of the survey. The results reported here are based on a prelim-
inary survey made in response to requests from school principals at three elementary
schools in Oklahoma. The three schools represented somewhat different demo-
graphic and socio- economic status areas from the state of Oklahoma. School A was
a rural school located in a small Oklahoma town , but because of its proximity to a
metropolitan area , drew on a population which included a high proportion of fami-
lies in which both parents were career professionals; School B was located in a rela-
tively poor , primarily Caucasian neighborhood of a major metropolitan area in
Oklahoma; and School C was another urban school from a different major metro-
politan area of Oklahoma. The principal of School C had invited all schools in her
district to send children of 3rd- grade age who were having problems in school (in
spite of intact auditory, visual , and general motor function); thus the set of School-
children represented a mixed and somewhat unusual population , with special needs.
Most of these children were from families which were lower socioeconomic status
(SES) and African- American.
In each case , we tested all the children in that school's third- grade population
who were judged by their teachers as having problems with reading. At one school
(School A) for a within- survey comparison , we also tested a subset of children judged
by the same teachers to be " high average " readers , three from each of the school's
six third- grade classes.
It should be emphasized that this survey was not designed as a rigorous experi-
ment , that is , with a formal experimental group vs. control group structure , nor was
there any effort to " match" groups according to features such as age , handedness
gender , SES etc. Also , subjects were accepted into the survey based on informal
teacher judgements of reading delay, rather than formal test scores. For this first step,
we were interested simply in determining how performance on the LAC vs. the
Visagraph II (each interpreted against the grade- level norms provided for that test)
would compare , in children who had previously been identified by their classroom
teachers as having difficulty with reading.
Subjects . As shown in Table a total of 131 children ranging in age from 7 11 to
3 participated in testing- 67 (49 reading- delayed plus 18 high- average) from
77235- 085- 096 10/20/03 9: 56 Page 87
School A , 27 from School B , and 37 from School C. As indicated in the table , there
were 15 instances of data drop- outs due to uninterpretable Visagraph II results
(excessive head movements during reading, interference from eyeglasses , etc. ), and
results from the final total of 116 children wil be reported here. Details comparing
age , gender and handedness of each of the groups are provided in Table I.
Procedures . Each school obtained informed consent from parents for this testing
to be done , with the understanding that the results would be made available both to
the parents and for inclusion in the child' s school record. At each school , two sepa-
rate and quiet testing rooms were identified , so that both types of tests could be done
in parallel- the LAC (along with a simple handedness questionnaire , both conducted
by JLL) and the Visagraph (conducted by PFM). Working in tandem in this way, we
were able to complete testing on approximately 20 children per each school day.
Results.
A. Phonemic awareness (LAC Test). Performance on the LAC test by these children
is summarized in Fig. 1. All 17 children in the high- average- reader group from School
A received scores that were above the recommended minimum for third grade; in
fact , LAC scores for 83% of those children were above the recommended minimum
for fourth grade. This result is in keeping with their reported good reading perform-
ance.
In the School A reading- delayed (RD) children , scores for 64% were below the
recommended minimum for third grade , 6% received scores in the 3rd grade- 4th
grade range , and scores for 30% were above the recommended minimum for fourth
grade. Thus for the latter 30% , something other than poor phonemic awareness must
account for their poor reading skils.
For the RD children from School B , 67% received LAC scores below the recom-
mended minimum for the beginning of third grade , 20% had LAC scores in the 3rd
... ,'"
normals (A)
school
FIGURE 1. Summary of performance on a test of phonemic awareness (the
Lindamood Auditory Conceptualization (LAC) Test) for children in four groups from
three schools: 17 high- average readers and 44 reading- delayed children from School
, 24 reading- delayed children from School B , and 31 reading- delayed children from
School C.
grade- 4th grade range , while scores for 13% of these children were above the rec-
ommended minimum for fourth grade. Finally, of the School- C RD children , 68%
had LAC scores below the recommended minimum , while 6% had scores in the 3rd
grade- 4th grade range , and 26% had scores above this.
Thus for all three schools (in spite of their ethnic and SES differences), approxi-
mately two- thirds of the reading- delayed children exhibited below- grade level
phonemic awareness abilities (cf. a similar proportion reported by Eden et al. 1994
1995), while one- third showed phonemic awareness abilities at or above this. Note
that some of these reading- delayed third- graders were even performing at a 7th-
grade (adult) level on this test. Thus phonemic awareness alone is clearly not suffi-
cient to account for reading problems in all these children- it may not even entirely
account for the reading difficulty experienced by the two- thirds with below- grade
phonemic awareness , since at least some of them may be additionally hampered by
undeveloped skils in another of the areas listed in Table III of the previous article
(Lauter , this volume).
085- 096 10/17/03 11:09 AM Page 89
177235-
B. Eye- movement coordination (Visagraph II). Visagraph scores for the same sets of
children are summarized in Fig. 2. It should be noted here that correlation coeffi-
cients comparing the two sets of grade level values were all below . 01 and non- sig-
nificant. The lack of correlation between the two sets of scores suggests that the two
tests are indeed independent , and assess different types of abilities- perhaps as dif-
ferent as predicted by the hemispheric specialization designations of Table III of the
previous article (Lauter , this volume). The lack of correlation also underlines the vari-
ety of individual differences presented by children in the classroom. Clearly neither
the concept that one curriculum fits all , nor the designation of a homogeneous phe-
nomenon of " dyslexia " can be intelligently defended in the face of such observa-
tions.
As shown in Fig. 2 , the above- average readers from School A included 18% with
eye- movement coordination scores below the third- grade level according to norms
for this test; 6% performed in the 3rd grade- 4th grade range , and 76% performed
above grade level. The RD children from School A included 80% with below- grade
performance , 9% were at grade level , and 11% were above grade level. For School B
58% of the RD children had eye- movement scores that were below grade level , 17%
were at grade level , and 25% were above grade level. Eye- movement scores for 100%
of the RD children from School C were below grade level. Note that for two of the
schools (A and C), even larger numbers of children exhibited below- grade- level eye-
100
for
Eye-movement coordination: summary
of children
::J)
,1.
normals (A)
school
movement scores than was true for phonemic awareness-yet eye movement coordi-
nation is not routinely tested in public schools.
C. Skill combinations-sub-populations?It is clear from Figs. 1 and 2 that for all these
groups of children , even the high- average readers , different combinations of skils are
represented. These combinations are summarized in Fig. 3 , presenting the percent-
age of children from each school who exhibited difrent combinations of grade- level
performance on the two tests.
The upper panel considers only those children in whom phonemic awareness
was below the recommended minimum. The graph plots the percentage of these
children from each school who had Visagraph scores which were: a) also below the
3rd- grade minimum; b) between 3rd grade and 4th grade; or c) above the 4th- grade
minimum. As the graph makes clear , in all three schools the most common combination
in RD children is below-grade- level perfrmance on both phonemic awareness and eye- move-
ment coordination: 51 % of the RD children from School A show this pattern , 42% from
School B , and 68% from School C. (The Trimodal Model would predict that this pat-
tern of performance should be found in at least some individuals of the " middle-
brain " type , characterized by blocked access to both right- and left- hemisphere skils.
The center panel makes the same comparisons for those children in whom
phonemic awareness was between the 3rd- and 4th- grade minimum recommended
scores. Few children are represented in these categories , since few of the RD children
had grade- level phonemic awareness.
The bottom panel ilustrates the combinations based on phonemic awareness
scores that were above the minimum recommended for the beginning of 4th grade.
This panel ilustrates that the next most common pattern for all three sets of reading- delayed
children is above-grade- level phonemic awareness combined with below-grade- level eye move-
ments. The prominence of such a pattern in these data may support the prediction of
the Trimodal Model that these particular children are of the " left-brain " type , char-
acterized by good access to left-brain skils such as phonemic awareness , but rela-
tively blocked access to right-brain abilities represented here by eye- movement
coordination , and also perhaps extending to poor mental imagery. Thus it may be
important to distinguish between children whose phonemic awareness is at grade
level (center panel) vs. those who have the skil in an advanced form (bottom panel).
As indicated by the data for the latter groups of children phonemic awareness that is
well above grade level may in some cases actually indicate a brain that has reduced access
right- brain specializations.
D. Importance of other skills?There is a third and somewhat more puzzling pattern
ilustrated in Fig. 3- at least some of the reading- delayed children (10% from School
A and 4% from School B) have both phonemic awareness and eye- movement coor-
dination scores that are at or above grade level. The obvious question is: " Why do these
children stil have reading problems?"
As noted in Table III of the previous article (Lauter , this volume), there are other
skils that predictably contribute to reading but were not formally assessed in this sur-
vey. For instance , mental imagery has been related both to performance on a design
reproduction test and to elementary- school math skils (Lindamood et al. , 1997). Our
085- 096 10/17/03 11:09 AM Page 91
177235-
nurmaJs (A)
;choul
Hannah (A)
school
Sf)
school
survey did not provide time for assessing imagery, but further work along these lines
might shed some light on how imagery interacts with these other skils to support
reading abilities.
These very preliminary findings suggest that by assessing a constellation of these
fundamental skils , we can characterize each and every child in a specific and gener-
ative way. Such an approach should be " generative " not only with regard to a better
understanding of the combination of abilities and lacks presented in each child , but
also an objective foundation for the design of customized training to ensure that all
children in a classroom are supplied with essentially the same toolbox of fundamen-
tal skils , with clear importance for all areas of learning, not reading alone.
exactly what program of training is best to help any particular individual become a
star lineman or gymnast or basketball player.
There are already a number of programs which have been used successfully,
some of them for many years , to improve fundamental skils for reading and learn-
ing in a variety of types of individuals. Examples include: 1) for phonemic awareness
the Lindamood Phoneme Sequencing (LiPS) program (formerly Auditory
Discrimination in Depth- cf. Lindamood et al. , 1997); 2) for mental imagery, the
Visualizing/Verbalizing Program (V IV - cf. Lindamood et al. , 1997); 3) for eye move-
ment coordination , the training programs used by developmental optometrists
known as Visual Training (cf. descriptions by Heath et al. , 1976 , Solan 1985 , Atzmon
1985; Cohen , 1988; Sigler & Wylie , 1994; Major & Kudija , 1994; Ciuffreda , 2002);
and the computer- based programs offered by Taylor Associates , based on the
Visagraph II; and 4) for general sensorimotor training, which may " tune up " a sys-
tem to better respond to these other types of training, exercises for whole- body motor
control and multisensory integration such as the MI- 2000 program (cf. Fadigan.
1998) and NeuroNet (designed by Nancy Rowe , cf. www. neuronetonline. com).
Each of these programs is available in versions appropriate for classroom use
and research has been done documenting effectiveness. For instance , one study using
a classroom version of Visual Training (the protocol recommended by the COVD
Optometric Guidelines for School Consultancy- d. Hellerstein et al. , 2001) was
recently conducted with children from one of the three schools (School C) mentioned
above (McKane et al. , 2001). Twenty- nine children with reading problems were
recruited and divided into an experimental and a control group. During three and a
half months of the spring semester , both groups received a combination of programs
designed to improve their auditory-verbal performance (including the LiPS and V /V
programs- d. Lindamood et al. , 1997). In addition , the experimental group was
treated with the COVD program implemented by a board- certified optometrist and
administered by school personnel , on a daily basis for 30 min/day. In a comparison
of measures taken at the beginning and end of the semester , both groups showed sig-
nificant improvement in reading scores on the Wide Range Achievement Test
(WRAT), but eye movements during reading showed significant improvement only
in the group which received the Visual Training.
Of course , behavioral changes with training need to be correlated with informa-
tion about the underlying biology. In order to identify individually- specific charac-
teristics related to the Trimodal Model , and to document changes in the central
nervous system associated with training, it is not necessary to use high- cost brain
imaging technologies such as functional magnetic resonance imaging (fMRI) or
positron emission tomography (PET). In fact, a combination of relatively inexpen-
sive techniques such as otoacoustic emissions , auditory brainstem response testing,
and small- array quantitative EEG (cf. Lauter , 1992), combined in a coordinated test-
ing design, such as employed in the Auditory Cross- Section (AXS) Test Battery
(Lauter , 1998b , 1999c , 2000a , 2002a , 2003) can serve not only to generate " neuro-
logical fingerprints " of individuals , characterizing each person in terms of dynamic
relations along all three body/brain axes (Lauter 1998b , 2000a), but also to document
, " , "
REFERENCES
Atzmon , D (1985). Positive effect of improving relative fusion vergences on reading and learn-
ing disabilities. Binocular Vision , 39- 43.
Ciuffreda , KJ. (2002). The scientific basis for and effcacy of optometric vision therapy in non-
strabismic accommodative and vergence disorders. Optometry, , 735- 762.
, j.
085- 096 10/20/03 9: 56 Page 95
177235-
Cohen , A.H. (1988). Special report: The effcacy of optometric vision therapy. Journal of the
American Optometric Association , 95- 105.
Colby, D. , Laukkanen , H. , Yolton , RL. (1998). Use of the Taylor Visagraph II system to
evaluate eye movements made during reading. Journal of the American Optometric
Association , 22- 32.
Eden , G. , Stein , j.F. , Wood , H. , Wood , F.B. (1994). Differences in eye movements and
reading problems in dyslexic and normal children. Vision Research , 1345- 1358.
Eden , G. , Stein j.F. , Wood , M. , Wood , F.B. (1995). Verbal and visual problems in reading
disability. Journal of Learning Disabilities , 272- 290.
Fadigan (1998). Effectiveness of the Structure of Intellect (SOl) assessment and MI- 2000
training program used with 760 inner- city middle- school and high- school students.
Unpublished report available from Learning Point , 6355 Metro West Blvd. Suite 455
Orlando FL 32835.
Heath , EJ., Cook , P. , O' Dell , N. (1976). Eye exercises and reading efficiency. Journal of
Academic Therapy, , 435.
Hellerstein , L.F. , Danner , R , Maples , W. , Press , L. , Schnee beck , j. (2001). Optometric
guidelines for school consulting. Journal of Optometric Vision Development , 56- 75.
Lauter j.L. (1992). Processing asymmetries for complex sounds: Comparisons between behav-
ioral ear advantages and electrophysiological asymmetries based on quantitative elec-
troencephalography (qEEG). Brain and Cognition , 1-20.
Lauter j.L. (1998a). Neuroimaging and the Trimodal Brain: Applications for developmental
communication neuroscience. Folia Phoniatrica et Logopedia , 118- 145.
Lauter , j.L. (1998b).Neurophysiological self- control: Modulation in all things. Journal of
Communication Disorders , 543- 549.
Lauter , j.L. (1999a). Functional asymmetries and the Trimodal Brain: Applications to devel-
opmental disorders. Journal of Developmental and Learning Disorders , 181- 260.
Lauter , j.L. (1999b). The Handshaking Model of Brain Function: Notes toward a theory.
Medical Hypotheses , 435-445.
Lauter j.L. (1999c). The AXS Battery: Promise for Communication Neuroscience. Presented
to American Speech- Language- Hearing Association , San Francisco CA.
Lauter , j.L. (2000a). The AXS Battery: Meeting the challenge of individual differences in
human brain/behavior relations. Behavioral Research Methods, Instruments and Computers
, 180- 190.
Lauter , j.L. (2000b). The brain and reading: Neural bases for a new perspective on reading
skils assessment and training. Invited presentation to Idaho State Conference on
Reading and Learning, Boise ID.
Lauter j.L. (2001). Is reading ' unnatural' for 2/3 of our children?- Right brain , left brain , and
the " missing link" skils for reading and learning. Invited keynote presentation to
SouthWest and Rocky Mountain division of American Association for the Advancement
of Science , Denton TX.
Lauter j.L. (2002a). The Auditory Cross- Section (AXS) Test Battery: A new way to study affer-
enUefferent relations linking ear , brainstem , and cortex. Presented to Acoustical Society
of America , Pittsburgh P A. Abstract: Journal of the Acoustical Society of America 211 , 2338.
Lauter , j.L. (2002b). Is reading ' unnatural' for 2/3 of children and adults? Presented to
American Speech- Language- Hearing Association , Atlanta GA.
Lauter , j.L. (2003). Otoacoustic emissions and the ' set of the center: ' a new test battery and
new statistical treatment. Presented to Acoustical Society of America , Nashvile TN.
Abstract: Journal of the Acoustical Society of America 113 , 2198.
Lauter , j.L. , Wood , S., Lynch , 0. , Schoeffler , L. (1999). Behavioral and neurophysiological
effects of meclizine in young adults , measured with otoacoustic emissions , REPs/ ABR
qEEG , and a computerized test of eye- hand coordination. Perceptual and Motor Skills, 88
707- 732.
77235- 085- 096 10/23/03 4: 56 Page 96
Mailing Addresses:
An audio- visual overview of some of these issues (Trimodal Brain model , the AXS
Test Battery, applications to reading) is provided in: J.L. Lauter (2002) Neuroimaging:
How Understanding Individual Difrences Can Improve Your Clinical Practice. (3- hr edu-
cational video with manual) Rockvile , MD: American Speech- Language- Hearing
Association. Available from ASHA professional catalog (1- 888- 498- 6699), item
#0112397.
77235- 0 9 7 - 11 0 10/17/03 5: 22 Page 97
Abstract. In this article, the author discusses why using technology is a critical component to
human development, why it should be included in developmentally based educational programs
for children with autism and how to avoid common setbacks when including technology in such
programs. Also given is the case study of "Alice a child with severe autism who made great
strides after a new technological artifact was introduced to the culture that included Alice and
her caregivers.
Introduction
Many parents and caregivers of those with autism have noticed how technology-
often in the form of computers , mechanical toys and electronic games- has a power-
ful impact on their children. Often children are able to attend for longer periods of
time , are more inquisitive and more active. In some cases , a child with poor propri-
oception may have better motor control when controlling a cursor via a mouse than
by pointing with her own finger. Though , in such situations , technology can be quite
powerful , it may come at a cost , such as the child becoming lost in the virtual world
some technology affords , or using the tool or toy exclusively for self- stimulating
behavior , or stimming. For these reasons , technology often is not incorporated fully
into most education plans for children with autism. Additionally, some technologies
may stifle natural social interactions , further adding to the concern that social skils
may suffer at the hands of technology. Stil , as this article wil discuss , technology not
only has a critical role to play in the development of children , but also can support
other educational goals. This article is broken into two parts. The first section dis-
cusses why including technology in a developmentally based education program is
important and provides suggestions to ease the inclusion of technology in such a cur-
riculum. The second section is a case study that provides how including technology
in an education plan can lead to unanticipated but positive results.
lThis is a phenomena many parents and teachers have pointed out to me. It is possible that this " move-
ment by proxy " is easier for those with proprioceptive problems because attention is focused on moving
another object , such as a computer cursor , instead of on moving one s own body.
9 7 - 11 0 10/17/03 5: 22 Page 98
177235- 0
DANIEL R. GILLETTE
If one were to list some of the obstacles facing those with severe autism , on the
list would probably be the very things for which we most employ technology. Also
for those with autism many technological artifacts , due to their refined and pre-
dictable nature , may be more entertaining; easier to comprehend or study; represent
adequate solutions to problems; and easier to interact with than humans. But tech-
nology that is so provocative and powerful may also be difficult to compete with for
attention , causing caregivers to limit how much such tools are used. Additionally, it
seems that the more magical and powerful the tool , such as computers , the more
compelling it is to children with ASD , but that very power can be difficult to com-
prehend and harness in a therapeutic setting. Stil , with proper selection and plan-
ning, high technology, such as computers and video games , can play an important
role in psycho- educational plans for children with autism.
9 7 - 11 0 10/17/03 5: 22 Page 99
177235- 0
dom work through computers , but on them; the computer becomes the target of our
attention. For this reason , computers pull at our energy and attention the way a half-
closed door into an unknown room beckons us to explore. Additionally, entering into
the workroom of the computer environment, it becomes difficult to interact with
those back on the other side of the proverbial looking glass. This is exacerbated by
the fact that we do not fully enter the virtual work environment, but straddle the edge
between the real and virtual. Socially, this is problematic , because it requires one to
focus most of his attention on the virtual workspace while also needing to attend to
the social cues of others located in the tangible world.
Though much of the software written for computers is intended for use by a sin-
gle person , many educational and multimedia software is either neutral in this regard
or actually meant to be used in a collaborative way. Stil , even when using collabo-
ratively focused software , interpersonal interactions tend to falter when we gather
around a single computer. Though blame is often rightfully rested on the shoulders
of the software engineers and designers , the main impediments are rooted in the tan-
gible computing objects , the hardware.
If a teacher and child wish to have a collaborative experience that includes using
a computer , many physical problems stand in their way. The following lists the most
common problems I have observed when people try to use computers in a social
manner:
Desktop and laptop computers require that both parties physically orient
themselves in the same direction , looking toward the screen. This makes it dif-
ficult to cue to communicative gesture or judge attentional focus.
. Standard computers have one set of controls , such as keyboard and mouse
meaning that though both parties can passively experience the computer at
the same time , only one person can actively engage with the device at any
given moment. Also , since these tools tend to be awkward to reposition and
9 7 - 11 0 10/17/03 5: 22 Page 100
177235- 0
require a specific physical orientation for use (that of being positioned squarely
in front of the screen and controls), turn taking requires a dance that is diffi-
cult to master. As a result, turn taking ceases to be spontaneous and becomes
somewhat scripted. Turn taking occurs at the end of a software- defined inter-
action or when something goes wrong, not when the overarching social con-
text dictates.
designing the use of technology into the education program. When considering what
types of technology should be included in a developmentally based psycho- educa-
tional program , there are key areas that need to be addressed. Technology should not
be used as a one- size- fits all solution or seen as appropriate for all situations , which
means there are no hard and fast rules for deciding when and which technology
should be used. Some reasons for using technology as part of a developmentally
focused education program are:
. To provide improved access , both physical and cognitive , to the environment
(mobility aids , home automation , sensory aids), communication/social settings
(augmentative communication devices and computer access devices that allow
virtual communities), information (schedulers , computer access devices to the
Internet and media translators , such as text to speech systems or Braile print-
ers) and other technologies (alternative computer input devices).
. To create an environment where certain content and skils training can be
delivered more effectively, such as computer games that provide an opportu-
nity for the child to practice skils required for escaping a burning building (for
examples , see http://www. dotolearn. com/).
. To provide a highly consistent representation of information and skils train-
ing (recorded speech libraries that pronounce words exactly the same each
time they are uttered).
. To provide a fun and rewarding experience (video games , videos , toys and
music players).
. To train a specific skil so more complex tools may be used in the future.
. To track user actions for later analysis (computer administered proficiency
tests or electronic counters).
Often we use technology in ways unintended by their creators , which makes
sense , since much of technology can be best described as tools , and when working
with tools , one must work with what is available to get the job done , even if the avail-
able tools are technically inappropriate for the task at hand. Therefore , we cannot
and should not rely on the marketing of technology to fully inform how it should be
used. Instead , each piece of available technology should be considered in context of
how , where and by whom it might be used. Here are some questions to ponder to
assist in making well informed technology decisions:
. What are the goals of the program unit: Without having a clear main goal , one
cannot pick appropriate tools to assist in that goal.
. What are the skils being targeted: Technology should be chosen so it supports
instead of negating the educational goals.
. What methods wil be employed in the unit: Since using technology for its
own sake is seldom a goal in a socially based program , one should choose
technology in regard to how it supports other pieces of the puzzle. Even when
teaching a tech- specific skil , a social component should be involved.
. What technology is available or can be made available in the environment:
Does it make sense to make do with what is available or to seek out a new
9 7 - 11 0 10/17/03 5: 22 Page 102
177235- 0
tool? If seeking out a new tool , how close should its intended use be to the way
it wil be utilized in the program? Also , are the costs and familiarization
process reasonable in relation to the need?
. What is the operational models) of the technology to be used: Consider how
one must orient herself, both physically and attentionally, to use the technol-
ogy. Also , be aware of environmental limitations of the technology that might
be a factor , such as , resistance to liquid spils or being dropped. Additionally,
physical access (how does one physically control the device) should be probed
in regard to the motor and cognitive abilities of the child.
. Why is each piece of technology being introduced: Surprisingly, this may be
the least investigated question when introducing technology into an educa-
tional plan , but by clarifying such a motivation , a mindset is encouraged that
wil prevent many missteps.
To avoid using technology simply for its own sake , a contextual focus on tool use
instead of focusing first on the tools available , makes sense. To ilustrate , a contextual
focus would yield the statement: to build a chair , one could use a saw , a hammer
nails and glue. A tool- oriented focus would look like this: a hammer can be used to
put nails into wood , pull nails out of wood and to break plaster. When a tool-based
focus is applied , there is a tendency to use the tool in question even if such a tool is
not required or contributes to learning in a meaningful way. When a contextual
approach to tool selection is taken , it is much more likely that selected tools wil be
used in service of the main curricular goals , versus being presented simply due to
availability.
Alce: An Anecdote
Last year , I had the opportunity to witness how a new piece of technology led to
positive outcomes for a child with severe autism. The ways in which the technology
played its role was somewhat unexpected , impacting not only the child directly, but
also reshaping the culture the child was a part of. The tool was a prototype voice out-
put communication aid (VOCA) that my colleagues and I were designing for
Archimedes Access and Research and Technology, Inc. , made possible by a grant
from Cure Autism Now. The design team consisted of Neil Scott, James True , Nina
Paley, Heather Clare , Kevin Gil and myself. This early prototype , while promising,
was stil quite embryonic and had a tendency to be a little quirky and unpredictable.
The VOCA consisted of a picture- based software application that allowed for the col-
lection of icons whose corresponding words would be spoken aloud by the device.
The hardware was a somewhat dated pen computer (it was slow) that accepted
presses by a stylus or finger nail on the screen as its main form of control (this
replaced the need for a mouse). What follows is an anecdotal account of one of the
children , Alice , involved in a usability study of the prototype VOCA.
Alice (a pseudonym), was a six year old girl with severe , regressive autism. While
quick to smile and apparently a generally happy child , at the beginning of the study
Alice had almost no communicative skil , aside from tugging on sleeves or reaching
for people and objects. Alice s ability to visually discriminate objects was at an almost
non- existent level in typical educational settings; she had no functional pencil grip;
showed little to no awareness of cause and effect; had no interest in computers; and
attended to auditory stimuli but showed no consistent interest in visual stimuli.
Alice s skils building had slowed dramatically in the year leading up to the study,
even with an intensive , multifaceted program. Twelve weeks after the study began
Alice was choosing the stylus in her attempts to control the VOCA , even though she
had finger pointers fashioned out of plastic guitar fingerpicks and a grown fingernail
that allowed for more direct control; was experimenting and problem solving in her
environment; could visually discriminate between around 20 objects; chose to work
with a caregiver on the VOCA to request objects instead of simply grabbing them;
9 7 - 11 0 10/17/03 5: 22 Page 104
177235- 0
and most significantly, was not only making decipherable (by familiars) attempts at
speech , but doing so in a contextually appropriate manner. It was a startling change.
Though Alice was almost rejected for participation in the study because she did not
possess skils at the level deemed appropriate for inclusion , she taught me the most.
In the year since this study, my gaze turns to the intersection of how the VOCA
was made part of the social activities involving Alice and how it supported experi-
mentation in emerging skils in such a setting. While my colleagues and I would be
more than happy to take credit for Alice s improvements , there was no evidence to
suggest that the VOCA had by itself led to most of her improvement, aside from
some strides in motor skils. Alice had a change in medication a few months earlier
but that also was unlikely to account completely for the observed changes in her
behavior and abilities. It is clear to me that how the VOCA became a focus of social
interaction had as much to do with Alice s improvement (if not more) than any other
factor. To understand how the VOCA set the stage for a change in the culture sur-
rounding Alice , we first must look at how the tools were chosen and introduced.
Selection
The decision to use a tablet computer was never doubted by the VOCA design
team. There was some interest in using smaller pen-based computers , such as palm-
sized PDAs (personal digital assistants), but early design research suggested that such
a form factor was less than optimal if the ultimate goal was communication via the
device at a close to normal pace , since the smaller screens made it more difficult
find and manipulate icons. There were limitations with the tablets that were readily
available at the time and in the range of the project budget-limitations of speed , bat-
tery life and display quality-but the tablets were quite durable and in general fit the
bill. I learned through earlier study that social interaction was more natural when
using the tablet , versus the desktop computers used in previous studies of mock- ups
primarily due to the fact that the tablets can be positioned and shared like a pad of
paper or a book. Stil , for Alice , this software and a tablet computer posed some
problems.
Though the VOCA software being tested provided natural speech output
(recorded voice instead of computer synthesized), which was expected to be of value
to Alice , the interface for selecting and using words was almost exclusively graphical
which played upon some of Alice s greatest weaknesses. Also , to use the device , Alice
would need to be able to point at , tap and drag objects on the computer screen
another weak area. Though alternative access methods could have been employed
none seemed any more appropriate , and if Alice could learn the skils for operating
the VOCA , primarily pointing, then the door to most other forms of technology
would be opened (pointing, when combined with another action , is the most com-
mon method for interacting with technology, be it triggering a power switch , chang-
ing a channel on a remote control or typing). As part of a holistic skils program , the
VOCA would serve a role well beyond its intended design of facilitating communi-
9 7 - 11 0 10/17/03 5: 22 Page 105
177235- 0
cation; it was expected to be a medium for the acquisition of skils that could be gen-
eralized to other situations.
Alice had no real interest in computers and had a tendency not to attend to visual
stimuli , so it was unclear how Alice would take to the VOCA. When first introduced
to the device , Alice showed little interest and was easily distracted. Though she
seemed curious about the speech output of the device , it wasn t enough on its own to
hold Alice s attention. Alice actually did not attend to the device for a sustained
period until there was a group made up of myself, her parents and one of her aides
all playing with the tool and being social around it , sharing our discoveries and col-
laboratively exploring the capabilities of the machine. It was in this setting that Alice
approached the device and cocked an ear to the aural output. If the device had not
been put into a social context, Alice probably would not have spent enough time
with it to get hooked by the speech output she so much appreciated. Using the
VOCA in a social context became the key for Alice to learn the features and
intended purpose of the device.
In general , all of these skils were already part of Alice s education plan , though
paper and caregivers acted in place of the tablet. Historically, when working on these
skils , Alice would attend for a few minutes and then lose focus. Alice was somewhat
easy to involve in activities , but had difficulty staying on track. Alice seldom worked
for a reward , so activities needed to be intrinsically rewarding. Though her caregivers
made noble attempts to make her skil building sessions more engaging, there was lit-
tle in the physical world that Alice was interested in. It was expected that there would
be similar difficulties in teaching Alice how to work with the VOCA , but over time
Alice was able to attend more and more to sessions that included the device , and by
experimenting with the device , Alice made significant strides. For instance , Alice
began to put her head near the computer screen while dragging the stylus across the
screen , which caused the cursor to follow. Alice spent a good amount of time exper-
imenting and observing the cursor/stylus relationship. A few days after this experi-
9 7 - 11 0 10/17/03 5: 22 Page 106
177235- 0
mentation began , Alice s mother observed her problem solving in the environment
for the first memorable time. Alice was asked to bring a bunch of potatoes from one
part of the kitchen to the other. Instead of following her usual routine of moving one
at a time , she placed the potatoes in a bag and moved them en- mass. Was there a
direct link between experimentation with the VOCA and this behavior? It is hard to
say, but Alice s caregivers suggest that she began to do more and more physical prob-
lem solving after the VOCA was introduced to her.
The reasons why Alice was able or more wiling to attend to sessions that
involved the VOCA are not clear , but the following seemed to contribute to her
interest:
. The intrinsic value of the aural output ,both in that it produced sound and that
Alice had a skew to the auditory, and that the device spoke , which was some-
thing Alice very much wanted to do , as witnessed by her regular attempts to
produce words
. The device was not introduced at first as a communication tool but just as a
new interactive object; as a result, it was something to engage in play with and
explore
. Since the device was new to the entire culture that grew up around Alice , she
was not the only " true " investigator , which made the playful investigation
more of a sincere collaboration where Alice was as likely to show her care-
givers a feature as they were to model for her
. The device could not be rushed in its response and did not " time out" in inter-
actions , but simply stood ready for the next interaction , meaning that Alice
had as much time as she needed to prompt the device and the response of that
prompting was never truncated
I believe that though all of these points may have been significant , the intrinsic
payoff for Alice of the device speaking, as well as the true collaborative environment
were what helped Alice the most to attend to both the VOCA and her collaborators.
The speech output may also have driven home the idea that the device had a role in
communication , since that is how all of her caregivers communicated , and was also
one of the ways Alice communicated before autism stole her voice. At around the
fourth week of the VOCA being available , Alice started to stand next to the device
when she wanted to ask for something, even though she was yet able to operate the
device on her own or had been taught to start communication attempts by going to
the VOCA; Alice had figured out that the VOCA was a way to communicate.
One of the interesting outcomes in three of the five children who were involved in this study was that
afterward they were reported to be more proficient in the use of picture communication books than
before , even when previous attempts to use PECS had failed to lead to any noticeable improvements in
communication. Though the reason for such an outcome is rather elusive , it is my belief that for these chil-
dren the concept of using pictures to communicate didn t click when using picture- only strategies , but after
using the VOCA , which gave pictures speech , that connection became more clear. Additionally, the way
others responded to VOCA use may have been more socially appropriate since the speech output was
more natural , and the response of others reinforced the meaning and value of the VOCA.
77235- 0 9 7 - 11 0 10/20/03 10:03 AM Page 107
evant, and instead was simply an alert to action. Also , since Alice s responses were
usually delayed , subtle and non-verbal , caregivers had a tendency to query again or
move on when Alice did not respond in within the natural timing window that was
opened (Alice was observed to subtly and appropriately respond to queries 10-
seconds after a circle of communication was opened , but by that time , she often no
longer had the caregiver s attention). After spending time with Alice making attempts
to operate the VOCA , the interaction style of caregivers changed , sans-VOCA. Over
the course of the study, the caregivers that were involved adopted a similar commu-
nication style , characterized by the following: all utterances were calm and metered;
repeating of utterances stopped in the majority of interactions; and there seemed to
be a near infinite patience in waiting for Alice to respond.
The verbal style of the caregivers became that of the VOCA. The VOCA acted
the way it did because it was a machine and could only respond when asked- it had
no mechanism to evaluate whether or not its output had registered on the user and
therefore had to wait for the next input. I can only hypothesize what caused the
change in caregiver behavior , but I suspect that it was a combination of falling into
the rhythmic interaction style the device encouraged and seeing Alice s response to
prolonged pauses. The outgrowth of this change in interaction style was:
. Alice now was required to complete her turn in the social dance
. Alice was given the time and freedom she needed to complete her turn
. Caregiver speech was dramatically slowed , which may have allowed Alice to
better process the sounds or meaning of words
. The tone and prosody of caregiver speech was now either contextually appro-
priate or neutral , the first utterance no longer being made ambiguous by an
alerting tone
. By requiring the VOCA be used as part of Alice s spontaneous queries (often
for objects , such as food), more of her social interactions became circular
instead of simply " point and get"
It is my current belief that the strides Alice made in speech , gestural communi-
cation , problem solving and visual processing happened because of the more social
and consistent interaction with her caregivers. I suspect that the VOCA provided
verbal modeling that Alice was able to mimic and that the device was a platform on
which internally developing skils could be externalized , though many of the these
elements did occur at other times in Alice s lie-it took the confluence of social (care-
giver change) and environmental (the introduction of the VOCA) factors for Alice to
make the progress she made. It was also the curricular notion that the VOCA could
be used as a learning tool in a variety of domains , instead of seeing the device as sin-
gularly purposed for communication.
Conclusion
The example of Alice shows the significant role technology can play in a devel-
opmentally focused education program. If Alice and her caregivers had immediately
9 7 - 11 0 10/17/03 5: 22 Page 109
177235- 0
attempted using the VOCA for communication , not only would there have been pro-
found frustration due to not knowing how to use the tool , but many opportunities for
discreet skil building would also have been missed. More importantly, if the care-
givers and Alice had not collaboratively explored the device s features , and instead
simply took discreet turns with the tool , not only would their VOCA related skils
have developed at a slower rate , but the wonderful social effects witnessed would not
have played out , even if Alice did not require social activity to motivate her to play
with the tool. Not only did the social focus of the curriculum help motivate Alice , it
also provided a setting for the care team to further develop their skils and relation-
ship with her. As this case ilustrates , with proper planning and focus , incorporating
technology into a skil building program can encourage social growth instead of
impeding it.
Mailing Address:
Abstract. UT tested whether children with autism integrate information from multiple sen-
sory modalities in speech identifcation of spoken syllables. An expanded factorial design was
used in which inftrmation from the fllee and voice was presented either unimodally or
bimodally, and either consistent with one another or not. After training the children speechread-
ing to enhance the influence of visible speech from the face, we repeated the identifcation task.
Children behaved similarly in the two replications, except for a larger influence of the visible
speech after training in speechreading. The fuz; logical model of perception (FLMP) was con-
trasted with a single- channel model (SCM) because they represent comparable integration and
non integration models, respectively. The model descriptions revealed that the FLMP gave a sig-
nifcantly better description of performance than the SCM, supporting the interpretation that
children with autism integrate vocal and facial inftrmation in speech perception. Given these
positive findings, we propose multimodal environments for learning language.
The goal of the present research is to gain some understanding of speech perception
in individuals with autism. Autism is a pervasive developmental disorder , which
apparently has increased from affecting approximately 1 in every 500 children
(Cowley, 2000) to 1 in 300 (M. I.N. D. Institute , University of California , Davis
http://www. dds. ca. gov/ Autism/ Autism main. cfm). Although the etiology of autism
varies, individuals diagnosed with autism must exhibit a) delayed or deviant lan-
guage and communication , b) impaired social and reciprocal social interactions , and
3) restricted interests and repetitive behaviors (American Psychiatric Association
1994). The language and communicative deficits are particularly salient, with large
individual variations in the degree to which autistic children develop the fundamen-
tal lexical , semantic , syntactic , phonological , and pragmatic components of language
(Tager- Flusberg, 1999). For the roughly one- half of the autistic population who
develop some form of functional language (Tager- Flusberg, 2000; Lord , Rutter
LeCouteur , 1994; Prizant, 1983), the onset and rate at which the children pass
through linguistic milestones are often delayed (e. g. no single words by age 2 years
no communicative phrases by age 3; American Psychiatric Association , 1994). Given
their limitations in language processing, a better understanding of their speech per-
ception should be particularly valuable. We begin with our extant view of the psy-
111
111- 144 12/11/03 9: 38 Page 112
177235-
Speech Perception
We define speech perception as the process of imposing a meaningful perceptual
experience on an otherwise meaningless speech input. The empirical and theoretical
investigation of speech perception has blossomed into an active interdisciplinary
endeavor , including the fields of psychophysics , neurophysiology, sensory percep-
tion , psycholinguistics , linguistics , artificial intelligence , and sociolinguistics. In any
domain of perception , one goal is to determine the stimulus properties responsible
for perception and recognition of the objects in that domain. The study of speech
perception promises to be even more challenging than other domains of perception
because it crosses all of these disciplines.
Our perception and understanding of speech is a multimodal process , influenced
by what we hear (the sound of the speakers voice) and what we see of the face and
accompanying gestures (Massaro , 1998). Research has repeatedly shown that pairing
the auditory speech with visual speech from the face produces a percept that is more
accurate and less ambiguous relative to presenting either of these modalities alone
(Massaro , 1984; Summerfield and McGrath , 1984). Viewing the speaker s face
increases the intelligibility of what is being said , especially when the auditory infor-
mation is degraded by noise (Sumby & Pollack , 1954) or hearing loss (Erber , 1969).
For example , viewing the speaker s face can improve intelligibility of the spoken
message as much as 15 dB in the speech to noise ratio (Sumby & Pollack , 1954).
Viewing the speaker s face to augment the spoken message is not limited to situ-
ations in which the auditory input is degraded. Perhaps the most compelling demon-
stration of the impact of visible speech on perception of the spoken message is the
McGurk effect (McGurk & Macdonald , 1976). In this classic demonstration , partici-
pants were presented a film of a young woman saying /aga/ that was dubbed with
the sound /abal. The participants often reported hearing /ada/ , putatively a fusion of
the place of articulation features of /aga/ and the manner and voicing features of /bal
(we provide an alternative explanation after our theoretical framework is developed).
When the dubbing process was reversed (an auditory /agal dubbed onto /abal lip
movements) participants sometimes reported hearing /abga/ , a combination of the
two syllables. Similar results were found with /pa/ and /ka/. This McGurk effect pro-
vides evidence that speech perception is a bimodal process , influenced by both the
sight and sound of the speaker. A theoretical account of bimodal speech perception
must describe how each source of information is evaluated , whether or how the
sources are combined or integrated , and how classification decisions are made.
111- 144 12/11/03 9: 38 Page 113
177235-
In the course of our research , we have found that the Fuzzy Logical Model of
Perception (FLMP) to be a universal principle of perceptual cognitive performance
that accurately models human pattern recognition (Massaro , 1998). People are influ-
enced by multiple sources of information in a diverse set of situations. These sources
of information are often ambiguous and any particular source alone does not usually
specify completely the appropriate interpretation.
The three processes involved in perceptual recognition are ilustrated in Figure
and include evaluation , integration , and decision. These processes make use of pro-
totypes stored in long- term memory. The evaluation process transforms these sources
of information into psychological values , which are then integrated to give an over-
all degree of support for each speech alternative. The decision operation maps the
outputs of integration into some response alternative. The response can take the form
of a discrete decision or a rating of the degree to which the alternative is likely. The
assumptions central to the model are: 1) each source of information is evaluated to
determine the continuous degree to which that source specifies various alternatives
2) the sources of information are evaluated independently of one another , 3) the
sources are integrated to provide an overall continuous degree of support for each
alternative , and 4) perceptual identification and interpretation follows the relative
degree of support among the alternatives.
Although speech perception has traditionally been viewed as a unimodal process
it appears to be a prototypical case of multimodal perception. As described in the
introduction , experiments in face- to- face communication have revealed conclusively
that our perception and understanding are influenced by a speaker s face , as well as
the actual sound of the speech (Massaro , 1998). Research has shown that the results
are well- described by the FLMP , which is an optimal integration of the two sources
of information (Massaro , 1998). A perceiver s recognition of an auditory- visual sylla-
ble reflects the contribution of both sound and sight. For example , if the ambiguous
auditory sentence, My bab pop me poo brive is paired with the visible sentence My gag
kok me koo grive the perceiver is likely to hear My dad taught me to drive. Two ambigu-
ous sources of information are combined to create a meaningful interpretation
(Massaro , 1998).
individuals with autism are impaired in both their face processing (Dawson et. aI
2002; Rogers & Pennington , 1991; Wiliams et al. , 2001), and their ability to integrate
information across modalities (i. e. Bryson , 1970; de Gelder , Vrooman , & Van der
Heide , 1991; Lelord , Laffont , Jusseaume , & Stephant , 1973; Martineau , Garreau
Roux , & Lelord , 1987; Waterhouse , Fein , & Modahl , 1996). However , the exact
nature of these impairments has not been specified and the evidence is limited. For
example , children with autism tend to avoid the face to face contact with others
required by shared attention (Happe , 1996) and , therefore , would naturally have less
experience with visual information from the face. It follows that we would expect
autistic children to be less influenced by the face in bimodal speech perception but
it is possible that they integrate the two sources in the same manner as normally-
developing children.
It is therefore essential to distinguish between how much information is obtained
from a sensory input and how information from multiple inputs is processed. Given
the FLMP framework , we make a formal distinction between " information " and
information processing. " The sources of information from the auditory and visual
channels make contact with the perceiver at the evaluation stage of processing. The
reduction in uncertainty affected by each source is defined as information. In the
description given by the FLMP , for example , the degrees of support for each alter-
native from each modality correspond to information. These values represent how
informative each source of information is. Information processing, on the other hand
refers to how the sources of information are processed. In the FLMP , this processing
is described by the evaluation , integration , and decision stages (see Figure 1).
Fortunately, there is an ideal experimental paradigm and theoretical analysis that
allow us to distinguish information from information processing, and to determine
whether integration occurs. The experiment involves the independent manipulation
of two sources of information in an expanded factorial design. It allows an assessment
of the influence of the many potentially functional cues , and whether or how these
cues are combined to achieve speech perception (Massaro , 1998). This systematic
variation of the properties of the speech signal and quantitative tests of models of
speech perception test how different sources of information are evaluated and how
they are actually used.
The goal of the present investigation was to assess speech processing abilities in
children with autism. We asked whether any observed differences are the conse-
quence of information or information processing. This distinction allows us to deter-
mine whether any decrement in performance is the result of an impairment in
crossmodal integration (information processing) or the inability to discern and utilize
the auditory and visual information (information). To address these questions our
experimental design was carried out in three stages. We began with an identification
test, utilizing an expanded factorial design. Given that the children appeared to inte-
grate crossmodally but were influenced only somewhat by the visual speech , we car-
ried out a training regiment in which the children were trained in speechreading. We
then repeated the identification task. We predicted that this training would result in
an increase in the influence of visible speech in the identification experiment , and
that their crossmodal integration would not change.
111- 144 12/11/03 9: 38 Page 116
177235-
Method
Participants
Seven children diagnosed with autism , 1 female , 6 males , ranging in age from 7
to 11 (M= 9. 87; SD=1.6) years were recruited from two different day programs for
children with autism in Santa Cruz County. Prior to the start of our investigation , we
requested parent permission from all of the children enrolled in the two school pro-
grams. Out of 12 children , permissions were given for these 7 children. Six of the
children were native speakers of American English; one child was a native speaker
Spanish , but spoke English fluently. Four of the children attended a private school
and the other three attended a special day program at a local public school in the
Santa Cruz area. All 7 of the participants were familiar with the experimenter (the
junior author) prior to our investigation and were unaware of the goals of the study.
Six of the seven children were capable of speech. The Appendix gives a detailed
description of each child.
Stimuli
All stimuli were presented by our computer- animated talking head , Baldi. Baldi'
speech and emotion are generated by a parametrically controlled polygon topology
(Massaro , 1998). The advantage of using the talking head derives from its ability to
mimic natural speech , by incorporating co articulation and being trained by natural
speech measurements (Massaro , 1998; Ouni et al. , 2003). The stimuli were the con-
sonant- vowel (CV) syllables Ibil and Idil and the vowel (V) syllable iii. The syn-
thetic visible speech was controlled and aligned with the synthetic audible speech
(Black & Taylor , 1997). The duration of the test syllables was 472 ms for Ibi/ 385 ms
for iii and 448 ms for Idi/. The intensity of the syllables was 64.4 dB-A. It should
be noted that Idil and iii look very similar in visible speech.
Figure 2 gives a diagram of the expanded factorial design used in this experi-
ment. The synthetic auditory and visual stimuli were presented unimodally and
bimodally in an expanded factorial combination , giving a total of 15 conditions.
There were 3 auditory conditions , 3 visual conditions and 3 x 3 or 9 bimodal condi-
tions. Each of the 15 conditions was sampled randomly without replacement in a
block of trials.
The identification task was presented before and after training in speech reading.
In the Pre- training test , there were 2 blocks across 5 days for a total of 10 observa-
tions under each of the 15 conditions. In the Post- training test, there were 4 blocks
across 5 days for a total of 20 observations under each of the 15 conditions.
111- 144 12/11/03 9: 38 Page 11
177235-
Ibil Iii
Visual
Idil None
Ibil
Auditory Iii
Idil
None
FIGURE 2. Expanded factorial design used in the Pre- training and Post- training
identification task. The auditory and visual stimuli were /bi/ /iI and Idi/ presented
unimodally or bimodally.
All stimuli was developed on a 600 MHz Pentium III with 128 MB memory and
running a Gforce 256 AGP-V6800 DDR graphics board running Microsoft Windows
NT 4 and a Graphic Series view Sonic 20" monitor. The Pre- training task was run on
the machine just described , whereas Training and Post- training tasks were run on a
Toshiba Satellite 5005- S504 laptop which has a 1 GHz Pentium III with 512 MB
memory and Nvidia GeForce2Go graphics running Microsoft Windows 2000 pro-
fessional. The auditory speech was delivered via Harman/Kardon internal speakers
or Plantronics PC Headset model SR1. Each student had the option to respond with
either an external mouse (Logitech M- CAA42) or a touch screen (KEYTEC Magic
Touch). Each child used the same response method throughout the experiment. All
sessions occurred at a computer workstation located in each school during Pre- train-
ing. Both Training and Post- training sessions were conducted individually at the stu-
dent's desk.
Procedure
The children were tested individually at their school. They were instructed to lis-
ten and watch Baldi on the screen , and to identify the consonant of the syllable " that
the speaker said" as either B or D. The participants made their responses by clicking
111- 144 12/11/03 9: 38 Page 118
177235-
on a labeled area on the screen directly below the test window. The experiment was
participant driven: the computer waited for the child to make a response before pro-
ceeding to the next trial. The investigator sat to the left of the child during the dura-
tion of the experiment, redirecting the child' s attention to the task if the child became
distracted and to supply uninformative motivational rewards for responses through-
out the investigation.
Training in Speechreadng
Stimuli
Baldi was also used to generate the consonant- vowel (CV) syllables Idi/ , lvii
/zi/ or Ibil. Figure 3 shows a view of Baldi at the onset of the articulation for each
of the four syllables.
Procedure
varied between 0 (no auditory information) and 1 (the auditory speech at 64.4 dB- A).
Table 1 gives the auditory levels used at each stage of training.
The students were instructed to watch Baldi and indicate the syllable that was
spoken. A 200 ms beep sounded prior to the presentation of the test stimulus to indi-
cate the start of each trial. Following the test presentation , response buttons appeared
in the upper left hand corner of the screen. Responses were made by activating a but-
ton labeled B , D , V , or Z presented in a 2 by 2 configuration , using the mouse or
touch screen. Placement of the labels was randomized across trials. Immediately fol-
lowing the student's response the buttons were removed and the next trial began.
Feedback was given for correct responses in the form of " stickers " and verbal praise
given by Baldi.
Before each training session , a test session was presented. Each syllable was pre-
sented visually without sound in 3 blocks of trials , generating a total of 12 trials (3
observations for each of the four syllables). Following completion of the 12 test trials
an accuracy score was calculated. If the student attained 100% identification accuracy
during the assessment, the student was congratulated and the program automatically
exited. If the student did not reach this criterion of 100% , then the program pro-
gressed to training (see Figure 4). The criterion level during training was 80% , which
had to be met for the child to advance to the reduced level of auditory input.
Students were given the option to select the color of Baldi after every 3 blocks (12
training trials). The students completed 3 sessions per week , which lasted approxi-
mately 30 minutes each. The students were given a 3- minute break between training
sessions. A " choice board" would appear on the screen and the student selected from
a variety activities and/or food items. Upon completion of the break , the experi-
menter would resume the training session. Training occurred for approximately
weeks or until the student was able to identify all stimuli with 100% accuracy on the
assessment test (without sound) for 2 consecutive sessions.
39.
41.9
10% 42.
20% 48.
30% 50.
40% 52.
50% 54.4
100% 64.4
111- 144 12/11/03 9: 38 Page 120
177235-
An identification judgment for each stimulus was recorded. The mean observed
proportion of identifications was computed for each participant for the unimodal and
bimodal conditions by pooling across all 10 replications of each condition.
The proportion of /di/ responses for each of the trial types was computed for
each participant. The top panel of Figure 5 gives the observed (points) proportion of
/di/ judgments as a function of the auditory and visual stimuli in the unimodal and
bimodal conditions. The children were clearly influenced by both the auditory and
visual speech in both the unimodal and bimodal conditions. Six of the points are cir-
cled in the top panel of Figure 5. The top three points correspond to the conditions
111- 144 12/11/03 9: 38 Page 121
177235-
auditory /di/ , visual /di/ , and bimodal /di/. The outcome that the proportion of /di/
responses was higher in the consistent bimodal condition than in the two correspon-
ding unimodal conditions is strong evidence that the children were integrating the
auditory and visual speech (see Massaro , 1998). If only a single source of information
111- 144 12/11/03 9: 38 Page 122
177235-
were being used on bimodal trials , the proportion of judgments could not be more
extreme than either of the unimodal proportions. The more extreme judgments
could only result from some combination (integration) of the two modalities. The
same interpretation can be given for the bottom three points for the syllable Ibil (the
proportion of /bil judgments is simply one minus the proportion of Idil judgments
in this two alternative task).
These conclusions are supported by separate analyses of variance carried out on
the visual , auditory, and bimodal conditions. Under the unimodal conditions , there
was a significant effect for the auditory factor F(2 , 12 ) = 12. 277 , P ~ 0. 01 and the
visual factor F(2 , 12) = 33. 879 , P ~ . 01 for the auditory and visual conditions , respec-
tively. In the bimodal condition , there were significant main effects for both the audi-
tory factor , F(2 , 12) = 14. , P ~ . , and the visual factor , F(2 , 12) = 6. , P ~ . 01.
However , the interaction between the two variables in the auditory- visual condition
did not reach statistical significance , F(4 , 24) = 0. 941 , P = 0. 54.
Training in Speechreading
One possible explanation for the small visual effect is a difficulty in speechread-
ing. This hypothesis coincides with previous findings that for children in general , the
auditory input provides more information than the visual input (see Massaro , 1984;
Massaro et al. , 1986). Recall that Massaro (1984) found that children were not as pro-
ficient as adults in their abilities to accurately identify the visual information under
the unimodal visual conditions and the children were also less influenced by the
visual information in the bimodal conditions. Thus , the ability to process visible
speech (and auditory speech) must be accounted for in order to address the question
of whether individuals with autism integrate information from these two modalities.
Previous findings show an improvement in speechreading through training. Our
question was whether children with autism could be trained to speechread more
accurately. We developed and implemented a computer-based speech reading lesson
focused on the visible aspects of speech using the consonant-vowel syllables /bi/
Idi/ , lvii and /zi/. These CV syllables were selected because they are reasonably
distinctive from one another and because /bil and Idil corresponded to the syllables
used in our experiment addressing the integration question.
Assessment data were captured daily for each student and continued until the stu-
dent was able to maintain 100% identification accuracy across 2 consecutive days of
assessments or 15 weeks of training. Other differences in the number of training ses-
sions across students are the result of absences and availability. Given these differ-
ences in the number of training sessions across students , we pooled the data across
sessions to give an equal number of training blocks across the 7 children. Student 6
reached the passing criterion (100% identification accuracy on the assessment for two
consecutive sessions) in 7 training sessions , and data for the remainder of the students
was pooled into 7 blocks.
--
111- 144 12/11/03 9: 38 Page 123
177235-
Figure 6 shows the average identification accuracy across the 7 blocks of trials.
As can be seen in the figure , identification accuracy improved systematically across
blocks. An analysis of variance with the proportion of correct identification as the
dependent variable and the block as the independent variable revealed that this
increase was significant , F(6 , 36) = 17.079 , P ~ . 01. Overall the students made
substantial gains from block 1 (M = . 37. , SD = . 09) to block 7 (M = . , SD = .14),
F(l , 6)= 30. 624 , P ~ . 01.
A second AN OVA , comparing the proportion of correct identifications in block
1 (M = . 37) to block 2 , revealed an 18% increase in accuracy and that this increase
was significant, F (1 , 6) = 12. 862 , P ~ 0. 01. This result indicates that the children made
substantial gains in speech reading performance after just one block of training.
assess gains after the initial block of training, we conducted an additional AN OVA in
which block 1 was eliminated from the analysis , revealing that the improvement
across blocks remained significant, F (5 , 30) = 13. 214 , P ~ 0. 01. These results indicate
that accuracy continued to increase as a function of training. Table 2 gives the indi-
vidual performance for each student. As can be seen in the table , each student
showed a substantial improvement in speechreading across the training sessions.
We then assessed accuracy for each syllable and its change as a function of train-
ing. As shown in Table 3 , accuracy increased for all syllables and the identification
of /bi/ and /vi/ was almost perfect by the end of training. The syllables /di/ and hi/
showed less improvement. These differences are reasonable because the syllable /bi/
0.5
0.4
-- 4 syllbles
3 syllbles
Block
FIGURE 6. Accuracy of identification in the training experiment across blocks. The
two curves correspond to accuracy computed for 3 and 4 alternatives , respectively.
111- 144 12/11/03 9: 38 Page 124
177235-
Table 2. Average proporton correct across the syllables /hi/, ldi/, lvii, and
Izil for each of the seven blocks of training.
Student
Block
0.42 0.17 0.43
0.48
0.44 0.42
0.42
1.00
1.00
0.41
0.40
0.49
can be distinguished by the closing of the lips and Ivil by the bottom lip tuck whereas
the syllables Idil and are very similar in visible speech except for duration (see
/zil
Figure 3). Given the similarity between Idil and /zi/ we also scored accuracy when
Idil and /zil were treated as one category. As shown in Figure 6 , this pooling
increased the overall level of performance , F(l , 6) = 37.890 , P ~ 0. , producing
almost perfect performance by Block
As in the Pre- training task , an identification judgment for each stimulus was
recorded and the mean observed proportion of Idil identifications was computed for
each participant for the unimodal and bimodal conditions by pooling across all 20
replications of each condition. The bottom panel of Figure 5 gives the observed
(points) proportion of Idil judgments as a function of the auditory and visual stimuli
in the unimodal and bimodal conditions. As in the Pre- training task , the children
were influenced by both the auditory and visual speech in both the unimodal and
bimodal conditions. One can also observe that the visual influence was much larger
111- 144 12/11/03 9: 38 Page 125
177235-
in the Post- training than in the Pre- training. The six points circled in the bottom
panel of Figure 5 ilustrate integration of the auditory and visual speech , following
the same logic we gave for the Pre- training task.
As in the Pre- training task , separate analyses of variance were carried out on the
auditory, visual , and bimodal conditions. Under the unimodal conditions , our results
revealed a significant effect for the auditory factor (2 , 12) = 19.410 , P ~ 0. 01 and the
visual factor F (2 , 12) = 188. 647 , P ~ 0. 01 for the auditory and visual conditions
respectively. Auditory-visual performance reveled significant main effects for both
the auditory factor and the visual factor , F(2 , 12) = 20. , P ~ . 001 , F(2 , 12) = 20.
P ~ . , respectively. However , the interaction between the two variables in the
bimodal condition did not reach statistical significance , F(4 , 24) = 1.091 , P = 0. 38.
we specify the prototypical features representing auditory /bil and dil in terms of
the onset of the second and third formants (F, ), whereas the prototypical features
representing visual /bil and Idil are given in terms of the amount of lip closure at
the onset of the syllable. Following this description , /bil and Idil are represented in
memory as
Idi/: falling F, F" and lips apart
/bi/: rising F, F" and lips closed
At the evaluation stage of processing, the input from each modality is evaluated inde-
pendently to determine to what extent it matches the prototype descriptions.
Independence means that the value assigned to one modality is independent of the
value assigned to the other. The degree of match is represented in terms of truth val-
ues in fuzzy logic , which can vary continuously between 0 (false) and 1 (true). For
example , an apple , a date , and an olive would be good , ambiguous , and relatively
poor members of the category fruit.
To ilustrate the predictions with just two response alternatives , rising F, F, can be
represented as (I- falling F, ) and lips closed as (I- lips apart). Assume that the audi-
tory input matches falling F, F, to degree . 8 and the visual input matches lips apart to
degree .4. Given just the auditory input, only the degree of match to the auditory fea-
ture would be relevant.
Idi/: falling F, F, = .
Given the relative goodness for response , the probability of a Idil response would be
81(. 8+. 2)
Analogously, given just the visual input, the probability of a dil response would be .4.
Given both of these auditory and visual inputs , then we have
Idi/: falling F, F, and lips apart = . 8 and .4
Idi/: 8 * .7 = .
/bi/: . 2 * . 3 = .
Using the relative goodness rule , the predicted probability of a Idil response is
P(!di/) = 561(. 56 06) = . 56/.62 = .
We see that the predicted probability of a Idil response is larger in the bimodal con-
dition than in either unimodal condition.
More generally, the FLMP is formalized in terms of the following equations. In
a two- alternative task with /bil and dil alternatives , the degree of auditory support
for can be represented by a. and the support for /bil by (1 - a.). Similarly, the
Idil
degree of visual support for dil can be represented by Vj' and the support for Ibil
by (1 - vJ The probability of a response to the unimodal stimulus is simply equal to
the feature value. For unimodal auditory trials , the predicted probability of a
response , P(! di/) is equal to
P(!dil) (1)
+(l a,)
For unimodal visual trials , the predicted probability of a response , P(!di/) is equal to
P(!dil) (2)
+(l
P(!dil) (3)
+(l a,) (1 -
These equations wil be implemented in the test of the FLMP against the current
results.
a bimodal auditory- visual speech event , the participant wil use the auditory infor-
mation with the probability p and the visual information with the probability (1 - p).
Given our earlier example with auditory and visual input and /bi/ and /di/ alterna-
tives , we can derive the probability of identification under the different conditions.
In the application of the SCM , it is assumed that the perceiver always uses the appro-
priate modality when only a single modality is presented. Assume that the probabil-
ity of a /di/ identification , P(!di/), given a specific auditory event is used is . 8. The
predicted P(!di/) is also . 8 on unimodal auditory trials because it is assumed that the
perceiver always uses the appropriate modality when only a single modality is pre-
sented. Similarly, assume that that P(!di/), given a specific visual event is used is .4.
The predicted P(!di/) is also .4 on unimodal visual trials because it is assumed that
the perceiver always uses the appropriate modality when only a single modality
presented. On bimodal trials , the response is determined by the probability of using
one modality rather than the other. The value of p varies between 0 and 1 and cor-
responds to the probability of using the auditory modality. The probability of using
the visual modality is simply 1 - p. If we assume p = . 7 in our example , the predicted
P(!di/) would be equal to the probability of using the auditory modality times the
probability of a /di/ identification of the auditory modality plus the probability of
using the visual modality times the probability of a /di/ identification of the visual
modality.
P(!di/) = . 7 * . 8 + . 3 * .4 = .
More generally,
P(!di/) = pa + (1 - p)V (4)
where p is the probability of using the auditory modality, a is the probability of a / di/
identification of the auditory modality, (1 - p) is the probability of using the visual
modality, and Vj is the probability of a /di/ identification of the visual modality. On
any trial , pattern recognition of a multimodal event is a consequence of only one of
the modalities (Massaro , 1987 , 1998).
Model Tests
To test auditory visual integration in speech perception , we tested the FLMP and
the SCM against each of the individual participant's results. As described in Massaro
(1998 , Chapter 2), the FLMP requires 6 free parameters: three parameters for each
auditory and visual stimulus to fit the 15 data points of the 3 x 3 expanded factorial
design. These parameters symbolize of the degree to which these modalities match a
prototypical /di/. The SCM requires 6 analogous parameters and a seventh corre-
sponding to the probability of using the auditory modality. The two models were fit
to the individual results and to the mean results across the seven participants.
Separate fits were carried out for the Pre- training and Post- training tasks.
The program STEPIT (Chandler , 1962) determined the quantitative predictions
of the models. Each model is represented as a set of unknown parameters and pre-
diction equations. STEPIT adjusts the parameter values of the model iteratively, min-
111- 144 12/11/03 9: 38 Page 129
177235-
imizing the root mean squared deviation (RMSD) between the predicted and
observed points. The RMSD provides an index of each model's goodness- of- fit
(Massaro , 1998).
In the Pre- training task , the RMSDs of the FLMP ranged from . 07 to .12 , with an
average RMSD of . 09. The fit of the mean participant gave an RSMD of . 03. The
RSMD of the SCM ranged from . 08 to .16 , an average of . , and the mean partici-
pant RMSD of . 05.
In the Post- training fit , the RMSD of the FLMP ranged from . 04 to . , with an
average of . 06. The fit of the mean participant gave an RMSD of . 04. The RMSD of
the SCM ranged from . 04 to .1 , gave an average of . , and the mean participant
RMSD of . 04.
The lines in Figure 5 give the average predictions of the FLMP. As can be seen
in the figure , the integration model is able to describe the results fairly accurately.
The three circled points in each of the panels show that the results and the model's
predictions both show a benefit of having consistent auditory and visual speech rel-
ative to either source presented alone.
Figure 7 gives the individual RMSDs for the FLMP plotted as a function of the
individual RMSDs for the SCM for both the Pre- training and Post- training tasks. The
points that fall below the diagonal line show a better fit of the FLMP over the SCM.
An analysis of variance was carried out on these RMSD values , with Pre- training and
Post- training and Model as independent variables. The FLMP gave a significantly
better fit of individual performance than did the SCM , F( 1 , 6) = 7.368 , P ~ . 05. Given
that the FLMP and SCM represent integration and nonintegration models , respec-
tively, we can tentatively conclude that the children were integrating auditory and
visual speech.
The RMSDs were significantly smaller in the Post- training than in the Pre- train-
ing task , F(1 6) = 8.135 , P ~ . 05. The reason for this difference is primarily due to hav-
ing twice than number of observations in Post- training than in Pre- training (Massaro
1998 , Chapter 10). Sampling variability decreases with increases in the number
observations. Although the fit of the models were better for Post- training than Pre-
training, the advantage of the FLMP did not interact with training, F(1 6) = 0. 362
P = . 57.
General Discussion
Our experiments provide some evidence that children with autism are influenced
to some extent by speech information in the face , can be taught to improve their sen-
sitivity to visible speech , and do integrate cross- modally in speech perception. We
tested an integration model (the FLMP) against a non- integration model (the SCM)
against the identification results from an expanded- factorial design in which the audi-
tory and visual speech were presented alone or together. Although the influence of
visible speech was relatively small in the first Pre- training test , we succeeded in train-
ing the children to speechread to allow a stronger test of integration when there was
111- 144 12/11/03 9: 38 Page 130
177235-
FIGURE 7. RMSD values for the fit of the FLMP as a function of the RMSD
values for the SCM for each of the seven children in the Pre- training (Experiment
and Post- training (Experiment 2) tasks. The diagonal line gives equivalent RMSD
values for the two models.
a larger influence of visible speech. The FLMP gave a significantly better fit than the
SCM across these two replications of the expanded- factorial design.
A possible limitation of our investigation is the absence of a control group. Even
though we did not test normally- developing children in the current study, however
there is an existing literature that makes such comparisons possible. Our question in
the present study was whether children with autism integrate auditory and visual
information in a speech perception task. Having now answered this question in the
affirmative , we can ask how this outcome compares with that of normally- develop-
ing children. Our previous research (Massaro , 1987 , Chapter 8) found that the FLMP
gave a significantly better description of performance than the SCM across a wide
range of development (3. 5 to 9. 5 years). Thus , the current results taken in conjunc-
111- 144 12/11/03 9: 38 Page 131
177235-
tion with previous research indicate that both autistic children and normally- devel-
oping children integrate auditory and visual speech as described by the FLMP.
Thus , we believe that we were able to address the integration question without a
control group. Our tests of the FLMP and SCM allowed us to determine whether
children with autism do integrate. The outcome of the model tests showed that the
FLMP better described the results than the SCM , a non- integration model. Given
this outcome , a working hypothesis is that children with autism do integrate the audi-
tory and visual speech in the bimodal condition. Future work should increase the
family of models that are tested , and to employ other measures of goodness- of- fit
(Massaro et al. 2001). Other non- integration models , such as an Auditory Dominance
Model (ADM), might adequately describe the results. We did not test the ADM in
the present study because it would have required almost as many free parameters (13)
as independent data points (15), making its test essentially invalid (Massaro , 1998). In
addition , RMSD is a relative measure of goodness- of- fit of a model , and it can be
supplemented with an absolute benchmark measure (Massaro , 1998). The bench-
mark measure provides a goodness of fit measure that should be expected if indeed
the model is correct. Other more complex techniques of model selection are being
developed and used (Myung & Pitt, 2003). Future work should employ these more
sophisticated techniques as much as possible to refine tests of the integration ques-
tion. We appreciate the fact that the results to date are stil not definitive. Given the
difficulty of testing children with autism in a task with many trials , the data are stil
somewhat limited. Also , the large spectrum of behaviors across the diagnosis of
autism makes it somewhat difficult to generalize across this population.
With respect to speechreading skil , we know that the overall performance accu-
racy would be lower for our children than for normally- developing age- matched con-
trols. We observed a fairly small influence of visible speech in the Pre- training
identification task. We know from our previous research that age- matched controls
would be more accurate in speechreading than were our children. Our children
whose average age was 9 years 8 months identified a visual /bil and Idil correctly
66% and 70% in the Pre- training task. In previous work, we found that normally-
developing fourth- grade children at the same age identified visual /bal and Idal 99%
and 97% of the time , respectively (Massaro , 1987 , Chapter 8). Thus we see that our
children began the study as much poorer speechreaders than their age- matched con-
trols. Consistent with previous training studies (Massaro , Cohen and Gesi , 1993;
Walden , Prosek , Montgomery, Scherr , &Jones , 1977), our autistic children became
better speechreaders with training. After training, the accuracy of identifying visible
speech of our 7 students improved significantly (90% and 82%) in the Post- training
task. This improvement stil fell a little short of that previously found for normally-
developing children at this age level (Massaro , 1987 , Chapter 8).
We also found in previous work that pre- schoolers , with an average age of 3.
years correctly identified visual /bal and Idal at rates of 81 % and 57%. Experienced
pre- schoolers , having had practiced identifying visible speech without feedback and
with an average age of 4. 75 years , correctly identified visual /bal and Ida! at rates of
90% and 82% (Massaro , 1987 , Chapter 8). Therefore , our children appear to be in
111- 144 12/11/03 9: 38 Page 132
177235-
Previous Literature
One of our conclusions is that children with autism , although somewhat limited
in their ability to use information from the face , integrate facial and vocal informa-
tion in speech perception. This conclusion appears to contrast with the previous lit-
erature. We review this literature and describe how the previous results either do not
address the issue of integration or can be more appropriately interpreted as a deficit
in processing facial information rather than a lack of integration.
Crossmodal influences have been studied using auditory evoked responses
(AERs) in a cross- modal association paradigm in which a sound is paired with a
strong visual stimulus , such as a flash of light (Martineau , Roux , Adrien , Garreau
Barthelemy, & Lelord , 1992). For instance , LeLord and colleagues (1973) discovered
that the occipital region of normal children showed an increase in amplitude of the
AER when a sound was preceded by a flash of light , whereas children with autism
showed no such increase.
However , the differences that were observed might simply be due to differences
in AERs without visual information rather than how the two modalities interact.
Martineau and colleagues (1987) found that the AERs for children with autism were
consistently smaller than children with normal intelligence but larger than children
with mental retardation. Martineau et al. (1992) found that children with autism
showed a great deal of variability in the AER and classified the performance into
three groups: below normal activation , activation comparable to controls , and above
normal activation. Moreover , the magnitude of the AER was positively correlated
with level of functioning. Although measurements of AER might eventually prove
informative , the research to date does not provide evidence for non- integration
across modalities.
Research supporting the hypothesis that the perceptual deficit results from inte-
gration difficulties is ostensibly supported in studies designed to measure associations
between auditory and visual information. Bryson (1970) found that the performance
of children with autism was lower when required to match auditory- to-visual and
visual- to-vocal events compared to auditory- to-vocal and visual- to- visual events.
Autistic children might perform poorer in this task simply because they have less
information about the visual and/or auditory events. More importantly, comparing
events across two modalities does not require an integration process. It simply
requires the perceiver to evaluate the relationship between the events from the two
modalities. Evidence suggesting that children with autism are able to match infor-
mation cross- modally (Boucher , Lewis , & Collis , 1998 , 2000; Walker- Andrews
Haviland , Huffman , & Toci , 1994) does not address the issue of integration. In the
111- 144 12/11/03 9: 38 Page 133
177235-
framework of the FLMP , the integration process is not necessary to perform the
matching task successfully.
De Gelder et al. (1996) compared bimodal speech perception of children with
autism to normal developing children matched for mental age. The investigators
used an expanded factorial design in which the children were presented with
bimodal , auditory- only, and visual- only speech segments. The children were pre-
sented with vowel- consonant- vowel speech segments: the consonants Ip/ , /b/ , It/
Id/ , Im/ or Inl paired with the vowel The children were instructed to watch and
Ia!.
listen to the speaker and identify the segments by repeating back what they thought
the speaker said. The influence of the visual speech on bimodal trials was measured
according to the proportion of " fused" or " combination " responses. Fused responses
were defined as those in which the place of articulation (visual speech) and the audi-
tory speech resulted in a single " fused" percept, such as identifying the combination
of a visual /bl and auditory Inl as Im/. Combination responses were defined as those
producing a combination of the syllables , such as identifying a visual /bl and audi-
tory Idl as /bdl. It appeared that the children with autism showed less influence of
the visual speech than the controls: the proportions of fused and blended responses
were .19 and . 51 for the children with and without autism , respectively.
The authors attributed this smaller visual influence on autistic children as an
inability to integrate the auditory and visual information. However , a smaller visual
effect found in the bimodal condition does not necessarily establish a lack of inte-
gration. Given the FLMP framework , we are able to make a distinction between
information " and " information processing. " The sources of information from the
auditory and visual channels make contact with the perceiver at the evaluation stage
of processing. The reduction in uncertainty produced by each source is defined as
information. In the fit of the FLMP , for example , the parameter values indicating the
degree of support for each alternative from each modality correspond to information.
These parameter values represent how informative each source of information is.
Information processing, on the other hand , refers to how the sources of information
are processed. In the FLMP, this processing is described by the evaluation , integra-
tion , and decision stages.
Given this framework , we are able to analyze the information and information-
processing differences between the two groups in the De Gelder et al. (1996) study.
Perceivers with hearing loss obviously have less auditory information , for example
but we can also ask whether they differ in terms of information processing. Similarly,
we can ask whether the integration process works the same way for children with
autism as for matched controls. According to the FLMP , the degree of influence is a
direct function of the information in the auditory and visual modalities. The propor-
tions of correct responses in the unimodal conditions for the children with autism
were . 97 and. 74 for the auditory and visual sources , respectively. We see that these
children were very good at identifying the auditory segments , but not so good at
identifying the visual segments. Thus , the FLMP would predict the contribution of
the auditory source of information would exert a greater influence than the visual
source in the autistic children s perceptual judgments of bimodal speech. The pro-
, "
portions of correct identifications for the control group were . 90 and . , respectively
for the unimodal auditory and visual conditions. These children were roughly
equally adept at identifying both the auditory and visual information. Thus , the
FLMP would predict that the two sources of information would influence perform-
ance in the bimodal condition roughly to the same degree. This analysis shows that
the FLMP can explain how observable behavior can differ without necessitating a
difference in the underlying information processing on bimodal trials. The FLMP
can predict that the children with autism were less influenced by the visible speech
than the matched controls , even though both groups of children integrated the two
sources of information in a similar fashion. When the results of de Gelder et al. (1996)
are analyzed in the framework of the FLMP , the outcomes for both the autistic chil-
dren and the control children are consistent with a process of integration of the audi-
tory and visible speech.
As pointed out by a reader of this article a potentially more parsimonious
account would say that autistic children are insensitive to emotions , social cues , and
face movements in part because they can t integrate them as well as normally devel-
oping children. " Lack of integration , however , should not necessarily lead to an
insensitivity of the cues within a given modality. Some individuals are required to
function unimodally (e. , the nonsighted in the perception of speech) and all of us
have unimodal inputs at least some of the time (e. , talking on the telephone). What
the results of the expanded factorial design make clear that small influence of visible
speech in the bimodal condition is a direct consequence of its small influence in the
unimodal condition. We carried out training in speechreading to enhance the impact
of the visual information and were successful in improving the children s speechread-
ing, which resulted in a larger impact of visible speech in bimodal speech perception.
Visual information from the face in speech perception is only one example of
many in which face perception plays an important role. Although the FLMP
describes speech perception , person identity and emotion may be processed analo-
gously (Schwarzer , & Massaro , 2001). While it is understood that the information nec-
essary for person identification and emotion differ from that for speech perception
recent research suggests that the information processes involved are identical (e.
Campbell , Schwarzer , & Massaro , 2001; Massaro , 1998 , Chapters 7 & 8). Thus , it
would be interesting to repeat the expanded- factorial design in tasks that manipulate
auditory and visual cues to these distinctions. These experiments would address the
question of integration abilities of autistic children in other domains that have also
been viewed as difficult for children with autism (Happe , 1996).
with autism (Bosseler & Massaro , in press). Our Language Wizard and Player
(described in Bosseler & Massaro , in press) encompass and implement developments
in the pedagogy of how language is learned , remembered , and used. Education
research has shown that children can be taught new word meanings by using dril
and practice methods. It has also been convincingly demonstrated that direct teach-
ing of vocabulary by computer software is possible and that an interactive multime-
dia environment is ideally suited for this learning (Berninger & Richards , 2002;
Wood , 2001). Following this logic , many aspects of our lessons enhance and reinforce
learning. For example , the existing program makes it possible for the students to
Observe the words being spoken by a realistic talking interlocutor (Baldi), 2) See the
word as spoken as well as written , 3) See visual images of referents of the words , 4)
Click on or point to the referent or its spelling, 5) Hear themselves say the word , fol-
lowed by a correct pronunciation , and 6) Spell the word by typing, and 7) Observe
and respond to the word used in context (see Figure 8).
Other benefits of our program include the ability to seamlessly meld spoken and
written language , provide a semblance of a game- playing experience while actually
learning, and to lead the child along a growth path that always bridges his or her cur-
rent " zone of proximal development." The Wizard allows the coach to exploit this
zone with individualized lessons , and with lessons that can bypass repetitive training
when student responses indicate that material is mastered.
The Bosseler and Massaro (in press) study consisted of two phases. Phase
measured vocabulary acquisition and retention. Phase 2 tested whether vocabulary
111- 144 12/18/03 4 : 06 Page 136
177235-
acquisition was due to the Language Player or outside sources and whether the
acquired words could be generalized to new images. Vocabulary lessons were con-
structed , consisting of vocabulary items selected from the curriculum of two schools
(Bosseler & Massaro , in press). The participants were eight children diagnosed with
autism , ranging in age from 7- 11 years- the same 7 children from the current study
and an eighth child. As noted earlier , all of the students exhibit delays in all areas of
academics , particularly in the areas of language and adaptive functioning, and seven
of the children were capable of speech.
The average results indicated that the children learned many new words , gram-
matical constructions and concepts , proving that the Language Player provided a
valuable learning environment for these children. In addition , a delayed test given
more than 30 days after the learning sessions took place showed that the children
retained many of the words they learned. This learning and retention of new vocabu-
lary, grammar , and language use is a significant accomplishment for autistic children.
Although all of the children demonstrated learning from initial assessment to
final reassessment , it is possible that the children were learning the words outside of
our learning program (for example , from speech therapists or in their school cur-
riculum). Furthermore , it is important to know whether the vocabulary knowledge
would generalize to new pictorial instances of the words , and outside of the learning
environment. To address these questions , a second investigation used the single sub-
ject multiple baseline design (Anderson & Kim , 2003). Once a student achieved
100% correct , generalization tests and training were carried out with novel images.
The placement of the images relative to one another was also random in each lesson.
Assessment and training continued until the student was able to accurately identify
at least 5 out of 6 vocabulary items across four unique sets of images. The students
identified significantly more words following implementation of training compared
to pre- training performance , showing that the program was responsible for learning.
Accuracy averaged about . , indicating that the learning extended to new images in
random locations. Most importantly, the children used this vocabulary knowledge
when tested outside of the Language Player environment by another teacher. These
results show that our learning program is effective for children with autism , as it is for
children with a hearing loss (Massaro & Light , in press).
In summary, although autistic children tend to be less influenced by the face , they
appear to integrate vocal and facial information in speech perception. These findings
in conjunction with previous evidence for integration in normally developing chil-
dren and adults , support a universal principle in which individuals optimally use
multiple sources of information in pattern recognition.
111- 144 12/11/03 9: 38 Page 137
177235-
Appendi
Diagnostic information for the eight children. The primary diagnosis for all of the
children is autism. The chart contains additional diagnoses , chronological age (C.A.),
non- verbal I.Q or measure of cognitive functioning (C. F.), level of adaptive func-
tioning, educational program, and reading level for each student.
9:11 N.A
rade
Mentally 11: 1 94** 38*** Beginning
retarded * First grade
9:4 38**** 57*** Beginning
Kindergarten
12:5
ten
*Information provided in next section
**Wechsler Intelligence Scale for Children-Third Edition (Wechsler , 1989).
***Vineland Adaptive Behavior Scales (Sparrow , Balla , & Cicchetti , 1984).
****Psychoeducational Profile Revised (PEP- R) Schopler , Reichler , Bashford , Lansing, &
Marcus (1990). This score represents the developmental age equivalent (in months).
g. "
111- 144 12/11/03 9: 38 Page 138
177235-
Student 1
Standardized Assessments:
Both verbal and non- verbal I.Q: could not be conclusively determined at last
testing. Clinician determined that he fell within the range of mental retardation.
His expressive language was so low that all other scores were altered.
Aggression:
Does not exhibit any observed aggressive behaviors toward others.
Student 2
Standardized Assessments:
See chart above
Aggression:
This child wil aggress towards self and others: pinching, hitting, screaming,
pulling hair , and scratching. Aggression is frequent (0- 5 times a day) and gener-
ally occurs to escape a difficult or undesired task.
111- 144 12/11/03 9: 38 Page 139
177235-
Student 3
Standardized Assessments:
Unable to obtain standard scores on either verbal or nonverbal standardized tests
due to non- compliance and disruptive behaviors.
Aggression:
Aggressive behaviors/tantrums are observed 0- 5 times a day, including kicking,
hitting, and head butting. These behaviors are typically observed when he is
required to complete an undesired task , unable to engage in desired activity, or
when leaving reinforcing items/activities.
Student 4
Standardized Assessments:
See chart above
Aggression:
Does not aggress toward others.
111- 144 12/11/03 9: 38 Page 140
177235-
Student 5
Standardized assessments:
See chart above
There is a large discrepancy in his I.Q; non- verbal I.Q is average (94), while his
verbal I.Q is 48 , placing him in the range of mentally retarded.
Aggressive Behavior:
Hitting, punching, screaming, scratching, and kicking. Aggression generally
occurs to escape a difficult or undesired task and occurs frequently (more than 5
aggressive episodes per day).
Student 6
Standardized Assessments:
See chart above
Aggression:
Engages frequently in aggressive behaviors (more than 5 aggressive episodes per
day). Displays non- compliance and aggressive behaviors directed toward others.
Aggression directed towards self and others , pinching, hitting, screaming, kick-
ing, and scratching generally occur during non- structured activities , new activi-
ties , difficult tasks , and transitions from one activity to another to escape/avoid a
non- desired activity transition and , during academic activities.
Student 7
Standardized Assessments:
According to the student's records , the severity of disability prohibits successful
participation in standardized testing.
Aggression:
Hitting, kicking, pushing, and pinching. Aggression is frequent (5- 10 episodes a
day) and typically occurs during transitions from one activity to another.
References
American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders
DSM-IV(4tl' ed. Washington , DC: Author.
Anderson , C. M. , & Kim , C. (2003). Evaluating treatment efficacy with single- case designs. In
M. C. Roberts & S. S. Ilardi (Eds. Handbook of Research Methods in Clinical Psychology
73- 91). Oxford , England: Blackwell.
Berninger , V. W. , & Richards , T. L. (2002). Brain literacy for educators and psychologists. San
Diego: Academic Press.
111- 144 12/11/03 9: 38 Page 142
177235-
Black , A. & Taylor , P. (1997). Festival speech synthesis system: System documentation (1.1.),
Human Communication Research Centre Technical Report HCRC/TR- Edinburgh.
Bosseler , A , & Massaro , D. W. (in press). Development and Evaluation of a Computer-
Animated Tutor for Vocabulary and Language Learning for Children with Autism.
Journal of Autism and Developmental Disorders. Under review. http://mambo. ucsc. edu/
pdf/autism.pdf
Boucher , j., Lewis , v. , & Collis , G. (1998). Familiar face and voice matching and recognition
in children with autism. Journal of Child Psychology and Psychiatry, , 171- 181.
Boucher , j., Lewis , v. , & Collis , G. (2000). Voice processing abilities in children with autism
children with specific language impairments , and young typically developing children.
Journal of Child Psychology and Psychiatry, , 847- 857.
Bryson , C. Q (1970). Systematic identification of perceptual disabilities in autistic children.
Perceptual and Motor Skill , 239-246.
Campbell , C. S. , Schwarzer , G. , & Massaro , D. W. (2001). Face perception: An information
processing framework. In: Michael j. Wenger , Ed; James T. Townsend , Ed.
Computational, geometric, and process perspectives on facial cognition: Contexts and challenges.
Lawrence Erlbaum Associates , Inc. , Publishers: Mahwah , NJ, p. 285-345.
Chandler j. P. (1969). Subroutine STEPIT: Finds local minima of a smooth function of sev-
eral parameter. Behavioral Science , 81- 82.
Cowley, G. (2000). Understanding Autism. Newsweek. July 31.
Dawson , G. , Carver , L. , Meltzoff, AN. , Panagiotides , H. , McPartland j., & Webb , S. j. (2002).
Neural correlates of face and object recognition in young children with autism spectrum
disorder , developmental delay, and typical development. Child Development , 700- 717.
de Gelder , B. , Vrooman , j., & Van der Heide , L. (1996). Face recognition and lip- reading in
autism. EuropeanJournal of Cognitive Psychology, 69- 86.
Erber , N. P. (1969). Interaction of audition and vision in the recognition of oral speech stim-
uli. Journal of Speech Hearing Research , 423-424.
Happe , F. (1998). Autism: An Introduction to Psychological Theory. Cambridge , MA: Harvard
University Press.
Kuhl , P. K. , &Meltzoff, A N. (1982). The Bimodal Perception of Speech in Infancy. Science
New Series , 218 , (4577). (Dec. 10 , 1982), pp. 1138- 1141.
LeLord , G. , Laffont, F. Jusseaume , P. , & Stephant j. L. (1973). Comparative study of condi-
tioning of averaged evoked responses by coupling sound and light in normal and autis-
tic children. Psycholophysiology, 415- 425.
Lord , C. ; Rutter , M. ; & Le Couteur , A (1994). Autism Diagnostic Interview- Revised: A
revised version of a diagnostic interview for caregivers of individuals with possible per-
vasive developmental disorders. Journal of Autism Developmental Disorders. , 659- 685
Martineau j., Garreau , B. , Roux , S. , & Lelord , G. (1987). Auditory evoked responses and their
modifications during conditioning paradigm in autistic children. Journal of Autism and
Developmental Disorders 525-539.
Martineau , j., Roux , S. , Adrien , j. L. , Garreau ,
B. , Barthelemy, C. , & Lelord , G. (1992).
Massaro , D. W. , Cohen , M. M. , & Gesi , A. T. (1993). Long- term training, transfer , and reten-
tion in learning to lipread. Perception Psychophysics , 549-562.
Massaro , D. , & Light j. (in press). Improving the vocabulary of children with hearing loss.
Volta Review in press.
Massaro , D. W Thompson , L. A , Barron , B. , & Lauren , E. (1986). Developmental changes in
visual and auditory contributions to speech perception. Journal of Experimental Child
Psychology 41(1): 93- 113.
McGurk , H. , & MacDonald j. (1976). Hearing lips and seeing voices. Nature 264 746-748.
1.N. D. Institute , University of California , Davis. http://www. dds. ca.gov/Autism/
Autism main. cfm
Myung, 1.j. & Pitt, M. A (2003). Model fitting. In L. Nadel (Ed. The Encyclopedia of Cognitive
Science , Vol. , pp. 47- 51. London , UK : Macmilan.
Guni , S. Massaro , D. , Cohen , M. , Young, K. &Jesse , A. (2003). Internationalization of a
Talking Head. 75th International Congress of Phonetic Sciences (ICPhS' 03), Barcelona , Spain.
Prizant, B. M. (1983). Language acquisition and communicative behavior in autism: Toward
an understanding of the " whole " of it.Journal of Speech Hearing Disorders. , 296-307.
Rogers , S. j. & Pennington , B. F. (1991). A theoretical approach to the deficits in infantile
autism. Developmental Psychopathology, , 137- 162.
Rosenblum , L.D. , Schmuckler , M. , & Johnson , j.A (1997). The McGurk effect in infants.
Perception Psychophysics , 347- 357.
Schwarzer , G. , & Massaro , D. W. (2001). Modeling face identification processing in children
and adults. Journal of Experimental Child Psychology, 79 (2), 139- 161.
Sumby, W H. & Pollack , 1. (1954). Visual contribution to speech intelligibility in noise. Journal
of the Acoustical Society of America , 212- 215.
Summerfield , A Q, & McGrath , M. (1984). Detection and resolution of audio-visual incom-
patibility in the perception of vowels. QJarterly Journal of Experimental Psychology, 36A
51- 74.
Tager- Flusberg, H. (1999). A psychological approach to understanding the social and language
impairments in autism. International Review of Psychiatry, , 355-334.
Tager- Flusberg, H (2000). Language development in children with autism. In L. Menn & N.
Bernstein Ratner (Ed. Methods For Studying Language Production , 313- 332). New
Jersey: Mahwah.
Van Lancker , D. , Cornelius , C. , & Needleman , R. (1991). Comprehension of Verbal Terms for
Emotions in Normal , Autistic , and Schizophrenic Children. Developmental
Neuropsychology, , 1- 18.
Walden , B. E. , Prosek , R. A , Montgomery, A. A , Scherr , C. K. , &Jones , C. j. (1977). Effects
of training on the visual recognition of consonants. Journal of Speech and Hearing Research
, 130- 145.
Walker- Andrews , AS. , Haviland j. M. , Huffman , L. , & Toci , L. (1994). Preferential looking
in intermodal perception by children with autism. Journal of Autism Developmental
Disorders , 99- 107.
Waterhouse , L. Fein , D. , & Modahl , C. (1996). Neurofunctional mechanisms in autism.
Psychological Review 103 , 457- 489
Wiliams j. H. , Whiten , A , Suddendorf, T. , & Perrett, D. 1. (2001). Imitation , mirror neurons
and autism. Neuroscience and Biobehavior Review , 287- 295.
Wood j. (2001). Can software support children s vocabulary development? Language Learning
& Technology, , 166-201.
111- 144 12/11/03 9: 38 Page 144
177235-
Mailing Address:
ACKNOWLEDGEMENT
The research and writing of the paper were supported by grants from National
Science Foundation (Grant No. CDA- 9726363 , Grant No. BCS- 9905176 , Grant No.
IIS- 0086107), Public Health Service (Grant No. PHS ROI DC00236), a Cure Autism
Now Foundation Innovative Technology Award , and the University of California
Santa Cruz. The authors would like to thank the teachers and staff at the schools and
the parents and children for their cooperation; Michael M. Cohen , Rashid Clark , and
Karl Young for their help at all stages of the research; and Justin Wiliams and Patricia
Lindamood for helpful comments on the paper.
,"
77235 145 146 10/24/03 8: 39 Page 145
Book Review:
A book dedicated to guiding the teaching of students with autism in the inclusive
classroom is long overdue. You re Going to Love this Kid!: Teaching Students with Autism
in the Inclusive Classroom is a landmark work by Paula Kluth , Ph. , an Assistant
Professor in the Department of Teaching and Leadership at Syracuse University. She
has been a special educator and inclusion facilitator and currently consults with
school districts across the country. In her book , she provides ready- to- use strategies
for including students with autism in both primary and secondary school classrooms.
First- person accounts of students autism give readers insight into the experience of
having autism and show educators how to adapt classrooms to support student par-
ticipation in class work , school routines , social activities and more.
This volume is unique in many ways in the world of teacher education. It focuses
exclusively on inclusive education as both ideology and pedagogy. Communication
behavior , and learning problems are understood in context and within relationships.
Useful strategies for teachers , administrators , therapists , counselors , etc. are included.
In addition , the voices of students with autism spectrum disorders are featured in a
sensitive and enlightening fashion. As the author notes ... these students are often a
catalyst for change and creativity. Specifically, including students with autism may
help teachers think more carefully about the choices offered to students: the design
of the lesson; the ways in which students can participate in teaching and learning;
and the comfort, engagement , and opportunities for all." (p. 31)
Kluth shows educators how to adapt their classrooms to support student partici-
pation in classwork , as well as school routines and social activities. The author skil-
fully weaves relevant research with lessons learned from her teaching experience to
give readers a comprehensive approach with specific ideas that are both pragmatic
and creative for:
values , and actions that support inclusive schooling
. understanding the attitudes ,
connecting, communicating, and collaborating effectively with families
145
77235 145 146 10/24/03 8: 39 Page 146
Mailing Address: